Language：English   Publishing type：Research paper (international conference proceedings)  

DOI： 10.1109/ICCE-Taiwan55306.2022.9869040

researchmap

Other Link： https://dblp.uni-trier.de/db/conf/icce-tw/icce-tw2022.html#NakaneAXFM22

Publishing type：Research paper (scientific journal)  

Current IaaS (infrastructure as a service) cloud service may not satisfy communication QoS (quality of service) requirements of delay-sensitive network applications, if there is a significant physical distance between a server of the network application (NetApp server) at a data center and its network application clients (NetApp clients). In order to improve communication QoS of NetApp clients, we propose server migration service (SMS) in this article. SMS allows NetApp servers to migrate among different locations in the network (1) to optimally locate themselves in relation to NetApp clients and mitigate the QoS degradation caused by location-related factors (ie, propagation delays on network links) and (2) to optimally distribute traffic load over routers and processing load over (physical) computers and decrease the energy consumption. We develop a mixed-integer programming model that determines when and to which locations NetApp servers migrate to minimize the total operating cost of SMS, that is, the sum of the monetary penalty incurred due to QoS violation and energy cost incurred due to energy consumption, while preventing NetApp servers from excessively migrating and adversely impacting QoS of the non-SMS service that share the resource of the substrate network with SMS. Simulation results show that the model developed in this article achieves up to 42% lower total operating cost of SMS compared to the model that only minimizes the monetary penalty of SMS without considering the energy cost of SMS.

DOI： 10.1002/ett.4511

Scopus

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)  

DOI： 10.1109/ICTC55196.2022.9952427

researchmap

Other Link： https://dblp.uni-trier.de/db/conf/ictc/ictc2022.html#DeguchiKTFY22

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

DOI： 10.1109/ictc52510.2021.9621088

researchmap

Authorship：Corresponding author   Language：English   Publishing type：Research paper (scientific journal)  

researchmap

Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

Edge computing is a promising computing paradigm that satisfies QoS requirements of delay-sensitive applications. In edge computing, server migration control is indispensable for managing client mobility. As a server migration method for edge computing, the method based on Q-learning has been proposed. However, the method assumes that there is only one application client and the number of destination edge servers is limited to one. In this paper, we propose a server migration method using Q-learning that copes with realistic situations where there are multiple application clients and destination edge servers. The contributions of this paper are as follows: 1) we clarify that, under the situation with multiple application clients and multiple destination edge servers, a straightforward server migration method using Q-learning (RL method) does not scale due to state space explosion, and 2) we propose a server migration method using Q-learning (RL-DR method) that reduces the dimensionality of state space by abstracting the numbers of application clients at all locations into a center of the gravity (COG) of application clients. The simulation results show that 1) RL method shows up to 248% worse performance than conventional server migration methods because of state space explosion and 2) RL-DR method achieves up to 38.3% better performance than the conventional methods.

DOI： 10.1109/ICOIN50884.2021.9333965

Scopus

researchmap

Other Link： https://dblp.uni-trier.de/db/conf/icoin/icoin2021.html#UrimotoFTMY21

Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

Various information is collected from IoT devices through the network. As such device becomes more familiar to the user, services are required to consider the influence of user. However, it is difficult to set the parameters of actuators that build consensus among all users in an environment where people with various preferences coexist. The conventional method minimizes the power consumption under the constraints of the user stress. However, this method has a problem that the calculation overhead is increased as the number of devices and users is increased. In this study, we propose a device control method based on consensus building with reinforcement learning. In the proposed method, the state is reduced by applying reinforcement learning for reducing the calculation overhead. As a result of evaluation, we clarified that our method obtains the device parameters that improve the reward by 1.5 times compared with the conventional method. Moreover, we also clarified that a reward value of 98.6% can be achieved compared to the optimum value.

DOI： 10.1109/ICOIN50884.2021.9333958

Scopus

researchmap

Other Link： https://dblp.uni-trier.de/db/conf/icoin/icoin2021.html#OishiTFY21

Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

Since network failures have a huge negative impact on our daily life, networks should be recovered from the failures as soon as possible. Although a lot of IP fast reroute methods have been proposed, most methods focus on multiple link failures or a single node failure. This paper proposes an IP fast reroute method which can reroute packets against multiple node failures. The paper is the first paper which deals with multiple node failures in the research area on IP fast reroute. The proposed method generates spanning trees to bypass the failures from a given network topology in network design stage, and reroutes a packet using one of the generated spanning trees every time the packet encounters a node failure in network operation stage. Numerical example shows that such spanning trees can be easily generated using our proposed method.

DOI： 10.1109/ICTC49870.2020.9289623

Scopus

researchmap

Other Link： https://dblp.uni-trier.de/db/conf/ictc/ictc2020.html#NumataITFY20

Publishing type：Research paper (international conference proceedings)  

© 2020 ACM. Recurrent neural network method considering traffic volume, timestamp and day of the week (RNN-VTD method) is a promising network traffic prediction method because of its high prediction accuracy. The RNN-VTD method encodes timestamp and day of the week, which are categorical data, to numerical data using label encoding. The label encoding, however, gives magnitude to the encoded values, which may cause misunderstanding of recurrent neural network models, and consequently, the prediction accuracy of the RNN-VTD method may be degraded. In this paper, we investigate the effect of using one-hot encoding instead of label encoding for a feature encoding method in the RNN-VTD method. In the one-hot encoding, each input data is encoded to an k-dimensional 0 - 1 vector where k is the number of category types. Because the encoded data do not have magnitude, it is expected that the prediction accuracy of the RNN-VTD method is improved.

DOI： 10.1145/3395245.3396441

Scopus

researchmap

Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

DOI： 10.1109/ICTC46691.2019.8939877

researchmap

Other Link： https://dblp.uni-trier.de/db/conf/ictc/ictc2019.html#KishimotoOTFY19

Language：English   Publishing type：Research paper (scientific journal)   Publisher：Springer New York LLC  

In this paper, we propose a new class of service called server migration service (SMS) to augment the existing IaaS (Infrastructure as a Service). SMS allows servers (server-side processes of a network application) to dynamically and automatically migrate as their clients (client-side processes of a network application) change their locations in order to reduce the total monetary penalty that the SMS provider pays to its SMS subscribers when failing to provide them with the guaranteed level of QoS. In this paper, we consider the monetary impact that arises from QoS degradation due to server migration and build an integer programming model to determine when and to which location servers should migrate to minimize the total monetary penalty incurred by the SMS provider. Numerical examples show that SMS achieves up to 96% lower total monetary penalty compared to that without server migration. Numerical examples also show that the integer programming model developed in this paper requires reasonable computation time under realistic parameter settings.

DOI： 10.1007/s10922-018-9451-6

Web of Science

Scopus

researchmap

Publisher：IEEE  

DOI： 10.1109/ICTC.2018.8539488

J-GLOBAL

researchmap

Publisher：IEEE  

DOI： 10.1109/NTMS.2018.8328725

researchmap

Publisher：IEEE  

DOI： 10.1109/ICTC.2018.8539630

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：Institute of Electrical and Electronics Engineers Inc.  

When a standard TCP implementation using the minimum retransmission timeout (RTOmin) of 200 ms is used in distributed file systems in data centers, a well-known throughput degradation called TCP Incast occurs, because 200 ms is too large as an RTOmin in data centers. In order to avoid TCP Incast, a TCP implementation using a much smaller RTOmin attained by a fine-grained kernel timer is proposed. However, throughput still decreases in the implementation when the number of servers in distributed file systems is large. In this paper, we propose an improvement method introducing a new TCP option which enables a client to request servers to retransmit packets in a better timing. Numerical examples obtained by extensive simulation runs show the effectiveness of the improved method.

DOI： 10.1109/ICTC.2017.8190962

Web of Science

Scopus

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：Institute of Electrical and Electronics Engineers Inc.  

The server migration service (SMS) is an optional service that improves communication QoS of the IaaS cloud service. In the SMS, small-scale data centers (micro data centers) called work places (WPs) are deployed at various locations in the network. In the SMS, a network application (a NW-App) consists of one or more server-side processes of the application (servers) and one or more client-side processes of the application (clients). In the SMS, servers migrate among WPs in order to improve communication QoS between the servers and their clients, unlike in the IaaS cloud service where locations of the servers are always fixed at a data center in the network. In our previous study, we developed an integer programming model and solved it to determine when and to which WPs servers should migrate in the SMS in order to minimize the financial penalty, i.e., SMS provider's financial loss due to degradation of the communication QoS. In this paper, we consider the electricity power cost, an important component of the operational cost of the SMS, in addition to the financial penalty considered in our previous study, in determining when and to which WPs servers should migrate in the SMS. In this paper, we define the operational cost of the SMS as the sum of the financial penalty and electricity power cost and consider a power-Aware server location decision problem in order to minimize the operational cost of the SMS. We formulate the problem as a mixed-integer programming model and solve the model numerically. Numerical examples show that our mixed-integer programming model optimally determines the locations of the servers and decreases the operational cost by up to 15.5% compared to our previous integer programming model, where the power cost was not considered.

DOI： 10.1109/ICTC.2016.7763457

Web of Science

Scopus

researchmap

Language：English   Publishing type：Research paper (scientific journal)   Publisher：Academy and Industry Research Collaboration Center (AIRCC)  

Minimum Physical Hop (MPH) has been proposed as a peer selection algorithm for decreasing inter-AS (Autonomous System) traffic volume in P2P live streaming. In MPH, a newly joining peer selects a peer whose physical hop count (i.e., the number of ASes traversed on the content delivery path) from it is the minimum as its providing peer. However, MPH shows high inter-AS traffic volume when the number of joining peers is large. In this paper, we propose IMPH that tries to further decrease the inter-AS traffic volume by distributing peers with one logical hop count (i.e., the number of peers or origin streaming servers (OSSes) traversed on the content delivery path from an OSS to the peer) to many ASes and encouraging the following peers to find their providing peers within the same AS. Numerical examples show that IMPH achieves at the maximum of 64% lower inter-AS traffic volume than MPH.

DOI： 10.5121/ijcnc.2016.8103

Scopus

researchmap

Language：English   Publishing type：Research paper (scientific journal)   Publisher：SPRINGER  

Light trail networks have been proposed as an extension of lightpath networks in order to enable sub-wavelength bandwidth allocation in WDM networks. The networks need a medium access control (MAC) method to avoid collisions of data transmissions. We have proposed a MAC method which adopts a token-passing mechanism. The method splits one trail into two trails: one upstream trail and one downstream trail and permits independent data transmissions on the trails. However, the method leaves bandwidths of links upstream from the token-holding node idle because the method only splits the original trail into two trails and leaves the upstream trail unsplit. In this paper, we propose a novel token-passing MAC method for further improving trail throughputs. Our proposed method recursively splits the upstream trail, and consequently the original trail can be split into more than two trails. In the proposed method, an appropriate setting of upstream/downstream token-holding times of all nodes is a key factor to successfully accommodate an input traffic. Therefore, we tackle a problem (called a token-holding time decision problem) where we optimize token-holding times of all nodes so that input traffic is successfully accommodated. We formulate the problem as a linear programming model. Numerical examples on maximum effective throughput show that the proposed method is more superior as the ratio of short-hop traffic is higher in input traffic patterns.

DOI： 10.1007/s11107-015-0509-z

Web of Science

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

In IaaS cloud services, QoS of network applications (NW-Apps) may degrade due to location factors such as significant distance between a server-side application (server) of a NW-App at a data center and a client-side application (client) of the NW-App at a client terminal. In order to shorten the distance and to improve the QoS, server migration services (SMSes) have been proposed. In SMSes, servers may migrate between different computers (called work places, WPs) on a network to prevent QoS degradation caused by the changes of client locations. Although server migrations can improve QoS of NW-Apps, they also generate a huge amount of traffic (server migration traffic) in the network. This paper focuses on a server location decision problem where the location of a server is decided in an on-line manner so that QoS of a NW-App is improved under the constraint that the server migration traffic has to be suppressed below an acceptable level. For the problem, we propose a practical on-line algorithm. The key idea behind the proposed algorithm is that the location of the server is decided with consideration of the QoS degradation in the future. The algorithm defines the averagely good location for the server where the QoS is expected to be relatively good for various client locations. Then, it keeps the range of the server's migration within the returnable range where the server can soon come back to the averagely good location. As a result, the QoS can be always kept as good as the one under the averagely good location. Simulation results show that the proposed algorithm improves QoS of the NW-App by up to 30% compared to a greedy algorithm.

DOI： 10.1109/CCNC.2015.7158070

Web of Science

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

When the standard TCP implementation is used in distributed. le systems in data center, a well-known throughput degradation called TCP Incast occurs. In order to avoid TCP Incast, a TCP implementation using a specialized. fine-grained timer is proposed. However, the throughput still decreases in the implementation when the number of servers in distributed. le systems is large. In this paper, we clarify that the cause of the decreasing is that maximum timeout value is too large and optimize the value.

DOI： 10.1109/ICTC.2014.6983174

Web of Science

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

The demand for high Internet availability has increased. As routing methods to recover fast from a single network component (link or node) failure, methods using multiple routing tables have been proposed. In the methods, when a component failure occurs, packets which are supposed to pass through the component in the failure-free state are rerouted using a routing table (backup table) which does not use the component for packet forwarding. Thus, it is important to decrease the increased routing cost (for example, the number of increased hops of rerouted packets if every link cost is one) due to the reroute as much as possible. In this paper, we propose a new routing method to decrease the increased routing cost. In the conventional methods, there are many components that do not fail but are not used for packet forwarding in a backup table. In our proposed method, we use such components as much as possible, and consequently we try to decrease the increased routing cost. Numerical examples show the effectiveness of our proposed method.

DOI： 10.1109/ICTC.2014.6983329

Web of Science

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

Split-connection type proxies have been proposed to improve TCP (Transmission Control Protocol) throughput in network environment with wireless LANs. A proxy is placed at the boundary of a wireless LAN (WLAN) and a wired network and it splits a TCP connection into two virtual TCP connections. We proposed a split-connection type proxy called MDA-Proxy (Modified Decreasing-ACK Proxy), which increases throughput of upflows (flows from wireless hosts to wired hosts) by decreasing the number of ACKs in a WLAN. MDA-Proxy uses two parameters, ack_stock and delay_timer, for each connection. The former controls the number of ACKs sent to a WLAN and the latter is used to recover from losses and delays of ACKs in a wired network. However, it is assumed that MDA-Proxy uses fixed values of the two parameters and their automatic tuning in response to network situation remains to be future work. In this paper, we propose a new type of proxy called ASA-Proxy (Ack_Stock Autotunning Proxy), which automatically tunes the two parameters in MDA-Proxy. Ack_stock is tuned comparing throughputs before and after its changing so that throughput of the connection becomes large, and delay_timer is replaced by the normal TCP retransmission timer because it is better than a fixed value of delay_timer for the inference of losses and delays of ACKs in a wired network. Extensive simulation runs show that ASA-Proxy is superior to MDA-Proxy.

Web of Science

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

In distributed file systems, a well-known congestion collapse called TCP Incast occurs because many servers send data to the same client and then many packets overflow the port buffer of the client link. Incast leads to throughput degradation in the network In our previous work, we have proposed a method to avoid Incast. In the method, we limit the maximum number of simultaneously existing connections to a predetermined constant value. However, we cannot use the method when the data size is small, and because we have not investigated how to optimize the maximum value, Incast may occur if the maximum value is not appropriate. In this paper, we improve the method so that it is applicable regardless of the data size and the maximum value is optimized Numerical results show the effectiveness of our proposed method.

DOI： 10.1109/ICTC.2013.6675396

Web of Science

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

In distributed file systems, a well-known congestion collapse called TCP Incast occurs because many servers send data to the same client and then many packets overflow the port buffer of the client link. Incast leads to throughput degradation in the network. In this paper, we propose two methods to avoid Incast based on the fact that only one connection can almost fully use the network bandwidth because the bandwidth-delay product is not so large in current data center networks. The first method is a method which completely serializes connection establishments. By the serialization, the number of packets in the port buffer becomes very small, which leads to Incast avoidance. The second method is a method which overlaps the slow start period of the next connection to the current established connection to improve throughput in the first method. Numerical results from extensive simulation runs show the effectiveness of our two proposed methods.

DOI： 10.1109/APCC.2013.6765931

Web of Science

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

The demands on Internet reliability and availability have increased. As a routing method to recover fast from a single network component (link or node) failure, a method using multiple routing tables has been proposed. In the method, when a component failure occurs, packets which are supposed to pass through the component in no failure state are rerouted using a routing table (backup table) which does not use the component for packet forwarding. Thus, it is important to decrease the increased routing cost (for example, the number of increased hops of rerouted packets if every link cost is one) due to the reroute as much as possible. In this paper, we propose a new routing method to decrease the increased routing cost. In the conventional method, there are such components that do not fail but are not used for packet forwarding in a backup table. In our proposed method, we use such components, and consequently we try to decrease the increased routing cost. Numerical examples show the effectiveness of our proposed method.

DOI： 10.1109/APCC.2013.6765984

Web of Science

researchmap

Language：English   Publishing type：Research paper (scientific journal)   Publisher：IEICE-INST ELECTRONICS INFORMATION COMMUNICATIONS ENG  

To improve TCP throughput even if the maximum receiving window size is small, a TCP performance enhancing proxy (PEP) using a UDP-like packet sending policy with error control has been proposed. The PEP operates on a router along a TCP connection. When the PEP receives a data packet from the source host, it transmits the packet to the destination host, copies the packet into the local buffer (PEP buffer) in case the packets need to be transmitted and sends a premature ACK acknowledging receipt of the packet to the source host. In the PEP, the number of prematurely acknowledged packets in the PEP buffer is limited to a fixed threshold (watermark) value to avoid network congestion. Although the watermark value should be adjusted to changes in the network conditions, watermark adjusting algorithms have not been investigated. In this paper, we propose a watermark adjusting algorithm the goal of which is to maximize the throughput of each connection as much as possible without excessively suppressing the throughputs of the other connections. In our proposed algorithm, a newly established connection uses the initial watermark value of zero to avoid drastic network congestion and increases the value as long as its throughput increases. In addition, when a new connection is established, every already-established connection halves its watermark value to allow the newly established connection to use some portion of the bandwidth and increases again as long as its throughput increases. We compare the proposed algorithm (CW method) with other methods: the FW method that uses a fixed large watermark value and the NP method that does not use the PER Numerical results with respect to throughput and fairness showed that the CW method is generally superior to the other two methods.

DOI： 10.1587/transcom.E95.B.2344

Web of Science

researchmap

Language：English   Publishing type：Research paper (scientific journal)   Publisher：IEICE-INST ELECTRONICS INFORMATION COMMUNICATIONS ENG  

Light trail architecture is attracting attention as a new optical wavelength-division multiplexing network architecture that can be built with currently available devices and can achieve bandwidth allocation with granularity finer than a wavelength. Because a light trail is a shared medium, we need a medium access control (MAC) protocol to prevent collisions. Although MAC protocols using token passing can prevent collisions, the bandwidths of links that are located upstream of the token holding node are kept idle. We first propose a dynamic light trail splitting method for increasing throughput of a light trail by using such idle bandwidths. Our method splits a trail into upstream and downstream trails at the token holding node, and independent data transmission on the two trails are permitted. As a result, we expect that the split trail architecture will achieve higher throughput than the original non-split trail architecture. The degree of performance improvement with the split trail architecture depends on how appropriately we determine the upstream and downstream token holding times of every transmission node. Thus, we formulate a problem in which we optimize the token holding times to accommodate requested traffic volume as a linear programming problem. We then derive the throughput of the split trail architecture by solving the problem using the NUOPT solver and investigate the degree of improvement over the original architecture. In addition, we evaluate the end-to-end delay of the split trail architecture by simulation. According to numerical examples, the split trail architecture achieves 1) almost the same throughput as the original one for the worst-case traffic pattern where every transmission node sends data to the terminating node of the trail only, 2) about 1.6 times higher throughput for a uniform traffic pattern where every node pair requests the same traffic volume and an extremely unbalanced traffic pattern where only a few node pairs request huge traffic volume, 3) about 1.9 time higher throughput for the split trail architecture's good-case traffic pattern where every transmission node sends data to its adjacent downstream node only, and 4) the end-to-end delay enough to satisfy any application's QoS requirement according to ITU-T Recommendation Y.1541.

DOI： 10.1587/transcom.E95.B.2330

Web of Science

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

In recent years, there is a big demand for live streaming service over the Internet. As one of the service model, peer-to-peer (P2P) live streaming has been utilized. In the system, it is important to decrease inter-AS (Autonomous System) traffic volume and content delivery delay because the larger volume leads to higher transit cost of ASes and this system treats the live contents, respectively. Therefore, as a conventional peer selection algorithm that tries to decrease them, Minimum Physical Hop (MPH) has been proposed. In MPH, to decrease inter-AS traffic volume, a newly joining peer selects a providing peer with the minimum physical hop count, which is the minimum number of ASes between the providing peer and the newly joining peer. Furthermore, to decrease content delivery delay, if there are several such candidates, it selects a providing peer with the minimum logical hop count, which is the number of Origin Streaming Servers (OSS) or peers passed from an OSS to the peer, from the candidates. However, MPH shows high inter-AS traffic volume when the number of joining peers is large. When we use MPH, candidates for providing peer with large logical hop count tend to swarm in ASes without an OSS while those with small logical hop count do in ASes with an OSS because of the peer selection policy of MPH. As the number of joining peers becomes large, the logical hop counts of the candidates for providing peer in every AS increase. As a result, in the former ASes, the logical hop counts of all the candidates for providing peer reach the upper bound frequently, and the newly joining peer cannot often find its providing peer in the AS, and consequently have to select providing peers in different ASes. In this paper, we improve MPH in order to further decrease the inter-AS traffic volume. In the improved algorithm called Improved Minimum Physical Hop (IMPH), we try to decrease the volume by distributing providing peers with one logical hop count to many ASes. As a result of the evaluation, we show that IMPH achieves to reduce the inter-AS traffic volume by at a maximum of 38% compared to MPH under uniform peer generation distributions.

DOI： 10.1109/ICNC.2012.31

Web of Science

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

TCP-fox is a new flow control method for TCP (Transmission Control Protocol) which allows the sender of a connection to send packets over the receive window advertised by the receiver to obtain higher throughput. However, TCP-fox has the problem that it halts sending packets when the receiver uses a receive buffer auto tuning mechanism. In this paper, we investigate the cause of the halting and present a way to avoid it.

DOI： 10.1109/APCC.2012.6388193

Web of Science

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：ACM  

Using network virtualization technologies for network applications (NW-Apps) consisting of server and clients, we can implement a server migration service where there are many server running environments (working places: WPs) inside a network and servers onWPs can migrate to otherWPs when communication QoSs in some NW-Apps fall off. Because server size is large, the traffic for server migration causes QoS degradation of its background traffic. Thus, it is important to decrease the degree (network impact) of the QoS degradation. In this paper, we propose three destination selection algorithms, MIA (Minimum Impact Algorithm), MRA (Maximum Remaining space Algorithm) and MCA (Maximum Covering Algorithm) that try to decrease the impact while increasing the number of NW-Apps' clients whose SLA (service level agreement) are satisfied. When a server migration is triggered, MIA, MRA and MCA move the server in the WP with the minimum impact, the WP with the maximum remaining space for servers and the WP with the maximum coverage, which is the number of routers under which the corresponding clients can communicate with the server while satisfying the SLA, respectively. Numerical examples show that when the number of accommodatable servers in full-cover WP (a full-cover WP is such WP that communicates with an arbitrary client while satisfying the SLA) is small or there is no full-cover WP, MIA outperforms the others, otherwise MCA outperforms the others. Copyright © 2012 by the Association for Computing Machinery, Inc.

DOI： 10.1145/2377310.2377315

Scopus

researchmap

Other Link： http://dblp.uni-trier.de/db/conf/cfi/cfi2012.html#conf/cfi/YamanakaFMYS12

Language：English   Publishing type：Research paper (scientific journal)   Publisher：ELSEVIER SCIENCE BV  

Change-point detection schemes, which represent one type of anomaly detection schemes, are a promising approach for detecting network anomalies, such as attacks and epidemics by unknown viruses and worms. These events are detected as change-points. However, the schemes generally also detect false-positive change-points caused by other events, such as improper parameter setting of detectors. Therefore there is a requirement for a scheme that detects only true-positive change-points caused by attacks and epidemics by unknown viruses and worms. The true-positive change-points tend to occur simultaneously and intensively in very large numbers, while the false-positive change-points tend to occur independently. Therefore, we expect that the multi-stage change-point detection scheme, which performs change-point detection in a distributed manner and takes account of the correlation among multiple change-points, can exclude false-positive change-points by neglecting those that occur independently. In this paper, we propose the multi-stage change-point detection scheme and introduce a weighting function that gives smaller weight to LDs with higher false-positive rate inferred by GD in order to avoid a set of false-positive alerts generated by the low-accuracy detectors from causing high false-positive rate of the scheme. We evaluate the performance of the scheme by a simulation using the parameter values obtained in an experiment using real random scan worms. In the evaluation, we modify AAWP (Analytical Active Worm Propagation) model so that it can derive the number of infected hosts (i.e., attack hosts) more accurately by considering a failure of infection behavior by random scan worms. The simulation results show that our scheme can achieve an optimal performance (detection rate of 1.0 and false-positive rate of 0) while the stand-alone change-point detection scheme, which does not use the correlation among multiple change-points, cannot attain such optimal performance, and our scheme with alert weighting always shows better detection performance than the scheme without alert weighting. (C) 2011 Elsevier B.V. All rights reserved.

DOI： 10.1016/j.comcom.2011.05.001

Web of Science

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

As a new optical WDM network architecture that can be built with currently available devices and can achieve bandwidth allocation with granularity finer than a wavelength, a light trail architecture attracts attention. Because a light trail is a shared medium, we need a medium access control (MAC) protocol to avoid collisions. Although MAC protocols using token passing can avoid collisions, bandwidths of links that locate upstream of the token holding node are kept idle. In this paper, we first propose a dynamic light trail splitting method in order to increase throughput of a light trail by using those idle bandwidths. Our method splits a trail into the upstream trail and the downstream trail at the token holding node and independent data transmissions on the two trails are permitted. As a result, we expect that the split trail architecture achieves higher maximum throughput than the original non-split trail architecture. The degree of throughput improvement by the split trail architecture depends on how appropriately we set upstream and downstream token holding times of every transmission node. Thus, we formulate a problem to decide the token holding times as a nonlinear programming problem, derive the maximum throughput of the split trail architecture by solving the problem using NUOPT solver, and investigate the degree of improvement compared to the original architecture. According to numerical examples, the split trail architecture achieves 1) almost the same maximum throughput as the original one for its unfavorite traffic pattern where every transmission node sends data to the terminating node of the trail only, 2) about 1.6 times as high maximum throughput for a uniform traffic pattern where every node-pair requests the same traffic volume, and 3) about 1.9 time as high maximum throughput for its favorite traffic pattern where every transmission node sends data to its adjacent downstream node only.

DOI： 10.1109/GLOCOM.2011.6133810

Web of Science

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

With the spread of broadband access lines, many bandwidth-consuming services such as video streaming have appeared. The popularity of these services can cause problems such as the shortage of the Internet backbone capacity, so it is important to detect increasing traffic caused by these services early. Conventional detection method called aggregation method tries to detect increasing traffic by predicting its future traffic volume with linear approximation. However, the aggregation method may not cope with increasing traffic whose growth is more rapid than linear growth. In this paper, we propose an early detection method (partial aggregation method) of increasing traffic under per-subnet based distributed traffic measurement. The method predicts 1) future traffic volume for each address in each subnet and 2) the number of subnets having increasing traffic in the future with a linear approximation. Then, the method estimates future traffic volume for each address as a product of the predicted future traffic volume in each subnet and the predicted subnet number. As a result, the method is expected to cope with rapid growth in traffic volume. Simulation results show that the partial aggregation method can detect increasing traffic earlier than the aggregation method by a maximum of 90 days.

DOI： 10.1109/TENCON.2010.5686580

Web of Science

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

As a detection method of large-scale simultaneous events (e.g., DDoS attack), a multi-stage change-point detection scheme with alert weighting was proposed. In the scheme, local detectors (LDs), which are deployed on each monitored subnet, try to detect an event by change-point detection. If they do, they send an alert to global detector (GD). Then GD judges whether an event is occurring by comparing the weight sum of the received alerts with an predetermined threshold value. The weight of an alert is set lower for LDs with higher false-positive rate (FPR). Conventional evaluation results only showed that alert weighting improves the performance for particular combination of two kinds of LDs with different FPRs. In this paper, we investigate the effectiveness of alert weighting for various combinations of two kinds of LDs with different FPRs in detail. We first consider the situation where detection rates (DRs) of all LDs are identical. Then, we consider the situation where high-FPR LDs show higher DR than low-FPR LDs, which is more realistic. Simulation results show that 1) alert weighting does not lead to degradation of detection performance and 2) alert weighting is most effective when event scale is moderate in our numerical examples.

DOI： 10.1109/TENCON.2010.5686586

Web of Science

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：ELECTRONICS TELECOMMUNICATIONS RESEARCH INST  

We evaluate performance of two medium access control methods with token passing, LT-FA MAC and ART MAC, in light trail networks. LT-FA MAC aims at coping with bursty packet arrivals while ART MAC aims at decreasing transmission delay. Simulation evaluation is performed in terms of the number of the maximum transmission nodes accommodated on a light trail and the maximum communication distance under the constraints that the bandwidth utilization rate of 90% and the average transmission delay of 100 ms must be achieved. Simulation results show that ART MAC permits about 500-1500km longer communication distance than LT-FA MAC with the same number of transmission nodes on a light trail and ART MAC accommodates about 15-25 more nodes than LT-FA MAC with the same communication distance.

Web of Science

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

One promising approach for large-scale simultaneous events (e.g., DDoS attacks and worm epidemics) is to use a multi-stage change-point detection scheme. The scheme adopts two-stage detection. In the first stage, local detectors (LDs), which are deployed on each monitored subnet, detects a change point in a monitored metric such as outgoing traffic rate. If an LD detects a change-point, it sends an alert to global detector (GD). In the second stage, GD checks whether the proportion of LDs that send alerts simultaneously is greater than or equal to a threshold value. If so, it judges that large-scale simultaneous events are occurring. In previous studies for the multi-stage change-point detection scheme, it is assumed that weight of each alert is identical. Under this assumption, false-positive rate of the scheme tends to be high when some LDs sends false-positive alerts frequently. In this paper, we weight alerts based on false-positive rate of each LD in order to decrease false-positive rate of the multi-stage change-point detection scheme. In our scheme, GD infers false-positive rate of each LD and gives lower weight to LDs with higher false-positive rate. Simulation results show that our proposed scheme can achieve lower false-positive rate than the scheme without alert weighting under the constraint that detection rate must be 1.0.

DOI： 10.1109/CQR.2009.5137356

Web of Science

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

In order to improve TCP performance in networks with large round trip time, a method using a PEP (Performance Enhancing Proxy) is proposed. In the method, the PEP is located on a intermediate router along a TCP connection, and returns acknowledgment packets (premature ACKs) to the source host instead of the destination host. In previous researches, although a congestion control in the PEP which keeps the number of prematurely acknowledged packets below a threshold (watermark) value has been proposed, packet losses are not taken into account, and consequently performance of the PEP under a packet-loss environment has not been investigated. In this paper, we incorporate a retransmission control to deal with packet losses into the congestion control. The proposed retransmission control mainly has two functions. One is to return duplicate premature ACKs or partial premature ACKs to urge the source host to execute the fast retransmission. The other is to retransmit prematurely acknowledged packets to the destination host, when a timeout occurs or three duplicate acknowledgments and a partial acknowledgment from the destination host are received. Simulation results under a packet-loss environment show that the PEP improves throughput larger as the end-to-end propagation delay becomes larger and throughput with PEP is a maximum of four times as much as throughput without PEP.

Web of Science

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

Change-point detection schemes, which represent one type of anomaly detection schemes, are a promising approach for detecting network anomalies, such as attacks and epidemics by unknown viruses and worms. These events are detected as change-points. However, they generally also detect false-positive change-points caused by other events, such as hardware problems. Therefore there is a requirement for a scheme that detects only true-positive change-points caused by attacks and epidemics by unknown viruses and worms. The true-positive change-points tend to occur simultaneously and intensively in very large numbers, while the false-positive change-points tend to occur independently. We can exclude false-positive change-points by excluding those that occur independently, based on information gathered from the entire network. In this paper, we combine change-point detection schemes with a distributed IDS, and evaluate performance of the combined scheme by a simulation using the parameter values obtained by an experiment using real worms. The simulation results show that the combined scheme detects all the DDoS attacks without any false-positives while we have to tolerate false-positive rate of at least 0.02 to detect all the attacks in a stand-alone IDS scheme.

DOI： 10.1109/APSITT.2008.4653540

Web of Science

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：IEEE Computer Society  

Change-point detection schemes are a promising approach for detecting network anomalies, such as attacks and epidemics by unknown viruses and worms. They detect those events as change-points. However, they generally also detect false-positive change-points, those caused by other events such as hardware trouble. A scheme is needed that only detects truepositive change-points, caused by attacks and epidemics. Truepositive change-points tend to occur simultaneously in very large numbers, while false-positive change-points tend to occur sporadically. We can exclude false-positive change-points by excluding change-points that occur sporadically, based on information gathered from the entire network. In this paper, we propose a multi-stage network event detection scheme that aggregates change-point information from distributed IDSs (Intrusion Detection Systems) and detects the true-positive change-points. Simulation results show that, compared to a scheme using only one IDS, our method always yields a smaller false-positive rate under the constraint that the detection rate of the true-positive change-points must exceed 0.99. © 2007 IEEE.

DOI： 10.1109/ICN.2007.71

Scopus

researchmap

Language：English   Publishing type：Research paper (scientific journal)   Publisher：SPRINGER  

We deal with the problem of designing the logical topology in IP-over-WDM networks. Many conventional methods for designing the logical topology assume that a constant number of wavelengths will be available on each fiber. But it is not necessary to utilize all wavelengths on each fiber in building an effective logical topology on a WDM network. Instead, several wave-bands may be considered for introduction while deploying additional wave-bands and their corresponding optical amplifiers when additional wavelengths are actually required. In this case, the number of wavelengths available on the respective fibers depends on the number of optical fiber amplifiers deployed on each fiber. In this paper, we propose a heuristic algorithm for the design of a logical topology with as few optical fiber amplifiers as possible. Our results indicate that our algorithm reduces the number of optical fiber amplifiers with a slight increase of average packet delays.

DOI： 10.1007/s11107-005-6019-7

Web of Science

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

Although the offset-time-based scheme can differentiate packet/burst loss rate according to their QoS levels in optical burst switching networks, it increases end-to-end packet delays of packets in a burst with higher QoS level. This is because it increases offset time of bursts in proportional to their QoS levels. In this paper, we propose a burst assembly method that decreases end-to-end packet delays. Our proposed method includes packets that arrive during the offset time into the currently assembled bursts while the conventional method includes them into the burst assembled next time. Simulation results show that our method offers more than 21% delay reduction for bursts with the highest QoS level.

DOI： 10.1109/ICCT.2006.341915

Web of Science

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

To improve TCP throughput performance, a method using a PEP (Performance Enhancing Proxy) has been proposed. The PEP operates on a router along a TCP connection. When a data packet arrives at the PEP, it forwards the packet to the destination host, transmits the corresponding ACK (premature ACK) to the source host in behalf of the destination host, and stores a copy of the packet into its own buffer (PEP buffer) in case retransmission of the packet is required. As a congestion control method on the PEP, a method which keeps the number of prematurely acknowledged packets in the PEP buffer below a threshold (watermark) value has been proposed. However, the relation between the watermark value and throughput is not sufficiently investigated, and an optimization method of the watermark value is not proposed. In this paper, we first investigate the relation between the watermark value and the average throughput. Extensive simulations show that the simulation results are roughly classified into two cases. In the first case, the average throughput becomes larger for larger watermark values and becomes a constant value when the watermark is over a certain value. In the second case, although the average throughput becomes larger for larger watermark values in the same way, it decreases when the watermark is over a certain value. Next, based on the results about the relation, we propose an watermark optimization algorithm which can adaptively maximize the average throughput of each connection and also satisfy a fairness condition that the average throughputs of connections are equal to each other.

DOI： 10.1109/ICCT.2006.341898

Web of Science

researchmap

Language：English   Publishing type：Research paper (scientific journal)   Publisher：IEICE-INST ELECTRONICS INFORMATION COMMUNICATIONS ENG  

The scalability of routing protocol has been considered as a key issue in large-scaled wavelength routed networks. Hierarchical routing scales well by yielding enormous reductions in routing table length, but it also increases path length. This increased path length in wavelength-routed networks leads to increased blocking probability because longer paths tend to have less free wavelength channels. However, if the routes assigned to longer paths have greater wavelength resources, we can expect that the blocking probability will not increase. In this paper, we propose a distributed node-clustering method that maximizes the number of lightpaths between nodes. The key idea behind our method is to construct node-clusters that have much greater wavelength resources from the ingress border nodes to the egress border nodes, which increases the wavelength resources on the routes of lightpaths between nodes. We evaluate the blocking probability for lightpath requests and the maximum table length in simulation experiments. We find that the method we propose significantly reduces the table length, while the blocking probability is almost the same as that without clustering.

DOI： 10.1093/ietcom/e88-b.10.3904

Web of Science

researchmap

Language：English   Publishing type：Research paper (international conference proceedings)  

researchmap

Publisher：Kluwer  

researchmap

J-GLOBAL

researchmap

J-GLOBAL

researchmap

J-GLOBAL

researchmap

J-GLOBAL

researchmap

J-GLOBAL

researchmap

J-GLOBAL

researchmap

J-GLOBAL

researchmap

J-GLOBAL

researchmap

J-GLOBAL

researchmap

J-GLOBAL

researchmap

J-GLOBAL

researchmap

J-GLOBAL

researchmap

J-GLOBAL

researchmap

J-GLOBAL

researchmap

Publisher：IEEE  

© 2018 IEEE. Server migration service (SMS) has been proposed as a new class of service to augment the existing IaaS (Infras-tructure as a Service) cloud service. SMS allows servers (server-side processes of a network application) to dynamically and automatically migrate close to their clients (client-side processes of the network application) in order to reduce the penalty that the SMS provider pays to is SMS subscribers when failing to provide them with the guaranteed level of QoS. In this paper, we tackle a link capacity provisioning and server location decision problem where we consider the sum of link capacity provisioning cost and the penalty as the total expenditure of SMS business, and aim at minimizing the total expenditure. In the problem, we determine how much capacity to add to links, and when and to which location to migrate servers. We formulate the problem as an integer programming model, solve the model, and obtain the optimal link capacities, server locations at each time, and the global optimum value of the total expenditure of SMS business. Numerical examples show that the proposed method decreases the total expenditure of SMS business by up to 49% compared to a conventional method that determines the server locations while fixing the link capacities.

DOI： 10.1109/CloudNet.2018.8549545

Scopus

researchmap

Other Link： https://dblp.uni-trier.de/db/conf/cloudnet/cloudnet2018.html#FukushimaYM18

Language：Japanese  

J-GLOBAL

researchmap

Language：Japanese  

J-GLOBAL

researchmap

Language：Japanese  

J-GLOBAL

researchmap

Language：Japanese  

J-GLOBAL

researchmap

Language：Japanese  

J-GLOBAL

researchmap

Language：Japanese  

J-GLOBAL

researchmap

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

CiNii Article

CiNii Books

J-GLOBAL

researchmap

Language：Japanese  

J-GLOBAL

researchmap

researchmap

researchmap

researchmap

researchmap

Language：English   Publisher：Academy and Industry Research Collaboration Center (AIRCC)  

In distributed file systems, a well-known congestion collapse called TCP incast (Incast briefly) occurs because many servers almost simultaneously send data to the same client and then many packets overflow the port buffer of the link connecting to the client. Incast leads to throughput degradation in the network. In this paper, we propose three methods to avoid Incast based on the fact that the bandwidth-delay product is small in current data center networks. The first method is a method which completely serializes connection establishments. By the serialization, the number of packets in the port buffer becomes very small, which leads to Incast avoidance. The second and third methods are methods which overlap the slow start period of the next connection with the current established connection to improve throughput in the first method. Numerical results from extensive simulation runs show the effectiveness of our three proposed methods.

DOI： 10.5121/ijcnc.2016.8406

Scopus

researchmap

Language：Japanese  

J-GLOBAL

researchmap

Language：Japanese  

J-GLOBAL

researchmap

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

Server migration service (SMS) is proposed as an optional service that can improve QoS of communications by network applications (NW-Apps) in IaaS cloud service. In SMS, computers called work places (WPs) are deployed at various locations in a network, and servers of NW-Apps can migrate to WPs where the servers attain better QoS. When a server migrates, it generates huge traffic, and consequently it causes QoS degradation (NW impact) of the background traffic on the server's migrating route. This paper tackles the server location decision problem where a server's location is decided so that the NW-App's QoS is kept as good as possible under the constraint that the degree of NW impact is suppressed below an acceptable level. For the problem, we propose a server location decision algorithm. The algorithm tries to improve the NW-App's QoS in the present and in the future by restricting the server's migration range within a neighborhood of the WP where the server is assumed to attain good QoS in many situations.

CiNii Article

CiNii Books

researchmap

Language：Japanese  

J-GLOBAL

researchmap

Language：Japanese  

J-GLOBAL

researchmap

Language：Japanese  

CiNii Article

CiNii Books

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

According to rapid increase of the Internet traffic, its total power consumption also increases. Therefore, it is necessary to decrease the total power consumption. Lightpath networks attract attention because they can relief the increase of the total power consumption by bypassing IP routers. Light trail networks have been proposed as an extended network architecture of lightpath networks. Light trail networks can achieve higher throughput than lightpath networks, and it is expected that light trail networks also show higher power efficiency than lightpath networks. In this research, we evaluate the power efficiency of light trail networks. In the evaluation, we reduce a power-efficient routing and wavelength assignment problem to a shortest path problem on an auxiliary graph, and solve it. In order to achieve this, we modify the conventional auxiliary graph for lightpath networks so that it can reflect the communication method of light trail networks. In addition, we calculate link costs of the auxiliary graph so that they corresponds to the increase of the total power consumption caused by the accommodation of traffic demands. Simulation results show that 1) light trail networks show higher power efficiency only when the network load is low and when the bandwidth granularity of communication channels is coarse, and 2) light trail networks show higher blocking probability than lightpath networks when they aim at minimizing the total power consumption.

CiNii Article

CiNii Books

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

Robust routing schemes are attracting attention as a routing scheme to achieve high robustness against changes in traffic patterns without performing both traffic measurement and dynamic change of traffic routes. This scheme, however, causes the performance degradation of TCP because it performs multipath routing and the out-of-order packet arrivals occur frequently. In this paper, we try to improve the performance by modifying a method (traffic distribution method) that determines how traffic should be distributed among the multiple routes for each source-destination (s-d) router pair that performs multipath routing. In the conventional traffic distribution method, the traffic distribution is determined on a per-packet basis. In this case, the number of packet delivery route changing is proportional to the number of the transferred packets, and consequently the out-of-order packet arrivals occur frequently. In this paper, we propose a per-burst distribution method. In this method, the traffic distribution is determined on a per-burst basis, where a burst is defined as a predetermined number of consecutive transferred packets for each s-d router pair. Numerical examples show that the per-burst distribution method can attain much higher throughput than the per-packet distribution method regardless of the burst length and can keep throughput high even when traffic patterns change if we use an appropriate burst length.

CiNii Article

CiNii Books

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

CiNii Article

CiNii Books

J-GLOBAL

researchmap

Language：Japanese  

J-GLOBAL

researchmap

Language：Japanese  

J-GLOBAL

researchmap

researchmap

researchmap

researchmap

researchmap

Language：Japanese  

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

In distributed file systems used in data center, a well-known congestion collapse called TCP Incast occurs because many servers send data to the same client simultaneously and then many packets overflow the port buffer of the client link. It leads to throughput degradation in the network. In this paper, we propose two methods to avoid Incast which called complete serialization and nearly complete serialization. The first method is a method which completely serializes connection establishments. By the serialization, the number of packets in the port buffer becomes very small, which leads to Incast avoidance. The second method is a method which overlaps the next connection to the current established connection to improve throughput in the first method. Numerical results from simulation with NS2 show the effectiveness of our two proposed methods.

CiNii Article

CiNii Books

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

Server migration service (SMS)is proposed to improve communication QoS of network applications (NW-Apps)operated in Infrastructure as a Service (IaaS). In SMS, work places (WPs), which refer to computers that run virtual servers of NW-Apps, are deployed at various locations in a network, and the virtual servers may migrate between WPs to achieve better communication QoS. In SMS, it is important to decrease the degree (impact) of negative effect on background traffic caused by migrating servers while keeping NW-Apps' communication QoS good. In this paper, under the assumption that multiple server can migrate at the same time, we propose two destination selection algorithms: EMIA (Extended MIA)and EMCA (Extended MCA)that take account of both NW-Apps' communication QoS and impact. In order to decrease accumulative impact, EMIA migrates servers to WPs with the minimum impact and EMCA migrates servers to WPs with the maximum coverage, which is defined as the range where NW-App can attain good communication QoS so that the number of server migrations needed in the future is decreased. Numerical examples show that the two algorithms can keep NW-Apps' communication QoS the same level and EMCA achieves smaller impact than EMIA.

CiNii Article

CiNii Books

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

In server migration service (SMS), computers called work places (WPs)are deployed at various locations in a network, and virtual servers for network applications (NW-Apps)may migrate between WPs to achieve better QoS. An SMS provider tries to provide an NW-App provider with a certain level of QoS that they agree upon, and if the SMS provider fails, it has to pay penalty (e.g., reimbursement of a part of service charges to users). Thus, the SMS provider is incentivized to migrate servers to satisfy the agreed-upon QoS level. 0n the other hand, an SMS provider also needs to be moderate in performing server migrations to avoid degradation of network QoS (i.e.,QoS of background traffic)caused by server migrations. This paper formulates an integer-programming model for the off-line server location decision (i.e., when and to which WP servers should migrate)where the penalty associated with NW-App's QoS violations is minimized, keeping the number and distance of server migrations below a given level. This paper also compares the minimum penalty obtained through solving the integer-programming model against the penalty obtained with a greedy on-line server locations decision algorithm, which migrates a server to a WP that minimizes the current penalty with no consideration of the penalty that will arise in the future. Numerical examples show that the integer-programming model achieves 36% to 49% lower penalty than the greedy algorithm when the degradation of network QoS is little acceptable.

CiNii Article

CiNii Books

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

CiNii Article

CiNii Books

J-GLOBAL

researchmap

Language：Japanese  

J-GLOBAL

researchmap

Language：Japanese  

J-GLOBAL

researchmap

Language：Japanese  

J-GLOBAL

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

Language：Japanese  

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

Robust routing schemes are attracting attention as a routing scheme to achieve high robustness against changes in traffic patterns without performing both traffic measurement and dynamic change of traffic routes. This scheme, however, causes the performance degradation of TCP because it performs multipath routing and the out-of-order packet arrivals occur frequently. As a routing scheme to improve TCP performance while keeping high robustness against changes in traffic patterns, we proposed a robust routing scheme called MDD (Minimum Delay Difference). For every node-pair, the MDD selects the multiple routes with the minimum propagation delay difference between them as the candidate routes, and then derives the traffic distribution ratio of them so as to minimize the worst case link utilization ratio. According to the numerical example of the conventional research, however, TCP throughput of the MDD is about 70% of its ideal value and there is room for further improvement. In this paper, for the purpose of further improvement of TCP throughput, we propose a hybrid robust routing scheme. The scheme only selects a single candidate route for a predetermined ratio of node-pairs in order to avoid TCP performance degradation cause by packet reordering while it selects multiple candidate routes as well as the conventional scheme for the rest of node-pairs. Numerical examples show that, under the condition that the worst case link utilization ratio must be less than 1.0, our proposed scheme achieves at the maximum of 24% higher throughput than the MDD.

CiNii Article

CiNii Books

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

Robust routing schemes are attracting attention as a routing scheme to achieve high robustness against changes in traffic patterns without performing both traffic measurement and dynamic change of traffic routes. This scheme, however, causes the performance degradation of TCP because it performs multipath routing and the out-of-order packet arrivals occur frequently. As a routing scheme to improve TCP performance while keeping high robustness against changes in traffic patterns, we proposed a robust routing scheme called MDD (Minimum Delay Difference). For every node-pair, the MDD selects the multiple routes with the minimum propagation delay difference between them as the candidate routes, and then derives the traffic distribution ratio of them so as to minimize the worst case link utilization ratio. According to the numerical example of the conventional research, however, TCP throughput of the MDD is about 70% of its ideal value and there is room for further improvement. In this paper, for the purpose of further improvement of TCP throughput, we propose a hybrid robust routing scheme. The scheme only selects a single candidate route for a predetermined ratio of node-pairs in order to avoid TCP performance degradation cause by packet reordering while it selects multiple candidate routes as well as the conventional scheme for the rest of node-pairs. Numerical examples show that, under the condition that the worst case link utilization ratio must be less than 1.0, our proposed scheme achieves at the maximum of 24% higher throughput than the MDD.

CiNii Article

CiNii Books

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

CiNii Article

CiNii Books

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

CiNii Article

CiNii Books

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

CiNii Article

CiNii Books

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

CiNii Article

CiNii Books

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

CiNii Article

CiNii Books

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

CiNii Article

CiNii Books

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

As a new optical network architecture that can be built with currently available devices and can achieve bandwidth allocation with granularity finer than a wavelength, the light-trail architecture attracts attention. For the light-trail architecture using token passing medium access control, the split trail architecture is proposed for improving the throughput. In the architecture, a trail is split into the upstream trail and the downstream trail at the token holding node and independent data transmissions on the two trails are permitted. However, the split trail architecture leaves the bandwidth of links upstream of the upstream-token holding node idle because the split of the trail is driven by the reception of only the downstream-token. In this paper, we propose the multi-split trail architecture for further improving the throughput. Our architecture utilizes the idle bandwidth by splitting the trail driven by the reception of both upstream- and downstream-tokens. We formulate the decision problem of the traffic accommodation where we optimize the token holding times so that the input traffic is successfully accommodated, as a linear programming problem. Then, we derive the maximum effective throughput of the multi-split trail architecture by solving the problem, and investigate the degree of improvement over the split trail architecture. According to numerical examples, the multi-split trail architecture achieves 1) the same maximum effective throughput as the split trail architecture for an uniform traffic pattern where every node pair has the same sending rate, and 2) about 1.93 times as high maximum effective throughput for its favorite traffic pattern where every transmission node sends data to its adjacent downstream node only and their rates are identical.

CiNii Article

CiNii Books

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

As a new optical network architecture that can be built with currently available devices and can achieve bandwidth allocation with granularity finer than a wavelength, the light-trail architecture attracts attention. For the light-trail architecture using token passing medium access control, the split trail architecture is proposed for improving the throughput. In the architecture, a trail is split into the upstream trail and the downstream trail at the token holding node and independent data transmissions on the two trails are permitted. However, the split trail architecture leaves the bandwidth of links upstream of the upstream-token holding node idle because the split of the trail is driven by the reception of only the downstream-token. In this paper, we propose the multi-split trail architecture for further improving the throughput. Our architecture utilizes the idle bandwidth by splitting the trail driven by the reception of both upstream- and downstream-tokens. We formulate the decision problem of the traffic accommodation where we optimize the token holding times so that the input traffic is successfully accommodated, as a linear programming problem. Then, we derive the maximum effective throughput of the multi-split trail architecture by solving the problem, and investigate the degree of improvement over the split trail architecture. According to numerical examples, the multi-split trail architecture achieves 1) the same maximum effective throughput as the split trail architecture for an uniform traffic pattern where every node pair has the same sending rate, and 2) about 1.93 times as high maximum effective throughput for its favorite traffic pattern where every transmission node sends data to its adjacent downstream node only and their rates are identical.

CiNii Article

CiNii Books

researchmap

researchmap

Language：English  

In a network environment including a wireless LAN and a wired network, split-connection type proxies have been proposed to improve TCP (Transmission Control Protocol) throughput. A proxy is placed at the boundary between a wireless LAN and a wired network, and it splits a TCP connection which is established between a host (WSTA) in a wireless LAN and a host (STA) in a wired network into two virtual TCP connections. We have proposed a split-connection proxy called DA-Proxy, and simulation results in our previous works show that although the DA-Proxy can increase throughput of downflows (flows from STAs to WSTAs), throughput of upflows (flows from WSTAs to STAs) is not improved and is almost the same as that of upflows without the proxy. In this paper, we propose a new proxy called MDA-Proxy which tries to increase throughput of upflows by decreasing the number of ACKs in a WLAN and show its effectiveness using simulation results. © 2012 IEEE.

DOI： 10.1109/GCCE.2012.6379960

Scopus

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

Server migration service provides network applications (NW Apps) with server running environment on a computer (WP) and service of server migration to another WP where the NW App can attain better QoS when it is degraded. Because the file size of a server is large, the traffic for server migration causes QoS degradation of its background traffic. Thus, it is important to decrease the degree (impact) of the QoS degradation. In this paper, we propose three destination selection algorithms MIA, MRA and MCA that try to decrease the impact while increasing the number of NW App's clients whose SLA are satisfied. When a server migration is triggered, MIA, MRA and MCA move the server to the WP with the minimum impact, the WP with the maximum remaining space for servers and the WP with the maximum communication range where the SLA is satisfied, respectively. Numerical examples show that when there is no full-cover WP (on which a server can communicate with an arbitrary client while satisfying the SLA) or its space is small, MIA outperforms the others, otherwise MCA outperforms the others.

CiNii Article

CiNii Books

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

Server migration service provides network applications (NW Apps) with server running environment on a computer (WP) and service of server migration to another WP where the NW App can attain better QoS when it is degraded. Because the file size of a server is large, the traffic for server migration causes QoS degradation of its background traffic. Thus, it is important to decrease the degree (impact) of the QoS degradation. In this paper, we propose three destination selection algorithms MIA, MRA and MCA that try to decrease the impact while increasing the number of NW App's clients whose SLA are satisfied. When a server migration is triggered, MIA, MRA and MCA move the server to the WP with the minimum impact, the WP with the maximum remaining space for servers and the WP with the maximum communication range where the SLA is satisfied, respectively. Numerical examples show that when there is no full-cover WP (on which a server can communicate with an arbitrary client while satisfying the SLA) or its space is small, MIA outperforms the others, otherwise MCA outperforms the others.

CiNii Article

CiNii Books

researchmap

Language：Japanese  

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

CiNii Article

CiNii Books

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

CiNii Article

CiNii Books

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

Server migration service is considered as one of IaaS-typed cloud services. The target of the service is such a application that consists of servers and their clients, and, each server works on a computer called a working place (WP) and can voluntarily migrate to another WP if necessary. However, if the destination WP is full of servers, a server in the WP may be forced to be pushed out of the current WP and migrate to another WP to accept the voluntary immigration server. In this paper, in order to protect some particular servers from experiencing voluntary migration or forced migration much more than the other servers, in the viewpoint of migration fairness, three methods to determine the candidates to be pushed out based on the numbers of voluntary migrations and forced migrations. Next, we propose two methods to determine such one pushed-out server from the candidates that the impact caused by forced migration against the network becomes as small as possible.

CiNii Article

CiNii Books

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

Server migration service is considered as one of IaaS-typed cloud services. The target of the service is such a application that consists of servers and their clients, and, each server works on a computer called a working place (WP) and can voluntarily migrate to another WP if necessary. However, if the destination WP is full of servers, a server in the WP may be forced to be pushed out of the current WP and migrate to another WP to accept the voluntary immigration server. In this paper, in order to protect some particular servers from experiencing voluntary migration or forced migration much more than the other servers, in the viewpoint of migration fairness, three methods to determine the candidates to be pushed out based on the numbers of voluntary migrations and forced migrations. Next, we propose two methods to determine such one pushed-out server from the candidates that the impact caused by forced migration against the network becomes as small as possible.

CiNii Article

CiNii Books

researchmap

Language：English   Publisher：The Institute of Electronics, Information and Communication Engineers  

As a new optical WDM network architecture that can be built with currently available devices and can achieve bandwidth allocation with granularity finer than a wavelength, a light trail architecture attracts attention. Because a light trail is a shared medium, we need a medium access control (MAC) protocol to avoid collisions. Although MAC protocols using token passing can avoid collisions, bandwidths of links that locate upstream of the token holding node are kept idle. In this paper, we first propose a dynamic light trail splitting method in order to increase throughput of a light trail by using those idle bandwidths. Our method splits a trail into the upstream trail and the downstream trail at the token holding node and independent data transmissions on the two trails are permitted. As a result, we expect that the split trail architecture achieves higher maximum throughput than the original non-split trail architecture. The degree of throughput improvement by the split trail architecture depends on how appropriately we set upstream and downstream token holding times of every transmission node. Thus, we formulate a problem to decide the token holding times as a nonlinear programming problem, derive the maximum throughput of the split trail architecture by solving the problem using NUOPT solver, and investigate the degree of improvement compared to the original architecture. According to numerical examples, the split trail architecture achieves 1) almost the same maximum throughput as the original one for its unfavorite traffic pattern where every transmission node sends data to the terminating node of the trail only, 2) about 1.6 times as high maximum throughput for a uniform traffic pattern where every node-pair requests the same traffic volume, and 3) about 1.9 time as high maximum throughput for its favorite traffic pattern where every transmission node sends data to its adjacent downstream node only. © 2011 IEEE.

DOI： 10.1109/GLOCOM.2011.6133810

Scopus

CiNii Article

researchmap

Language：English  

Fixed robust routing is a promising approach for ISP networks to accommodate variable traffic patterns with low operational complexity. The routing minimizes the maximum link load by distributing traffic of every source-destination pair to multiple paths (multipath routing). The multipath routing can result in performance degradation of TCP because of frequent out-of-order packet arrivals. In this paper, we first investigate the influences of delay differences among multiple paths and ratio of shorter paths among multiple paths on TCP performance with simulation using ns-2. The simulation results clarify that smaller delay difference and lower ratio of shorter paths among multiple paths lead to higher TCP throughput. Based on the investigation results, we next propose fixed robust routing algorithms that try to improve TCP throughput in addition to decreasing the maximum link load. The first algorithm called MDD (Minimum Delay Difference) selects a set of paths with the minimum delay differences between the shortest and the longest paths as candidate paths for every source-destination pair The second algorithm called MDD-LF (Minimum Delay Difference with Limited Fraction) bounds the fraction of traffic routed on the shortest delay path in addition to selecting the same candidate paths as MDD. Simulations using ns-2 show that, compared to a straightforward fixed robust routing that selects k-shortest hop paths as the candidate paths, MDD and MDD-LF achieve about 22% and 27% higher TCP throughput while MDD and MDD-LF produce about 1.7 and 2.3 times higher maximum link load, respectively. © 2011 IEEE.

DOI： 10.1109/CQR.2011.5996088

Scopus

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

As a new optical network architecture that can be built with currently available devices and can achieve bandwidth allocation with granularity finer than a wavelength, the light trail architecture attracts attention. For the light trail architecture using token passing medium access control, the split trail architecture is proposed. In the architecture, a trail is split into the upstream trail and the downstream trail at the token holding node and independent data transmissions on the two trails are permitted. As a result, we expect that the architecture achieves higher maximum throughput than the original unsplit trail architecture. The degree of throughput improvement by the split trail one depends on how appropriately we set upstream and downstream token holding times of each transmission node. In this paper, we formulate decision problem of the token holding times as a non-linear programming problem, derive the maximum throughput of the split trail architecture by solving the problem using NUOPT solver, and investigate the degree of improvement compared to the original one. According to numerical examples, the split trail architecture achieves 1) the same maximum throughput as the original one for its unfavorite traffic pattern where every transmission node sends data to the terminating node of the trail only, 2) about 1.6 times as high maximum throughput for an uniform traffic pattern where every node pair has the same sending rate, and 3) about 1.9 time as high maximum throughput for its favorite traffic pattern where every transmission node sends data to its adjacent downstream node only.

CiNii Article

CiNii Books

J-GLOBAL

researchmap

Language：Japanese  

J-GLOBAL

researchmap

Language：Japanese  

J-GLOBAL

researchmap

In recent years, peer-to-peer (P2P) live streaming systems are emerging because of their high scalability and robustness. In the systems, peer selection algorithms that determine logical topologies of P2P overlay networks have a great influence on important performance metrics such as the maximum number of joining peers and inter-AS (autonomous systems) traffic volume. As conventional algorithms that take account of those metrics, MLH (Minimum Logical Hop) and MPH (Minimum Physical Hop) have been proposed. In MLH, a newly joining peer (say Pnew) selects peers whose logical hop counts are as small as possible, where the logical hop count of a peer is defined as the number of hops from an origin streaming server (OSS) to itself. When there are too many such peers, Pnewselects peers whose physical hop counts are as small as possible, where the physical hop count between two peers is defined as the number of ASs between them. In MPH, Pnewselects peers in the reverse order of MLH. These conventional peer selection algorithms use physical hop count under the assumption that complete AS topology information is available. However, it is hard to obtain the complete information. In this paper, we modify the conventional algorithms so that they use each peer's belonging AS information only instead of AS topology information in order to make them practical. In the modified version of MLH and MPH (MLH' and MPH'), Pnewselects the peers within the same AS as Pnewinstead of peers whose physical hop counts are as small as possible. In addition, in order to distribute top level peers (i.e., peers that directly retrieve video from an OSS) to many ASs, MPH' gives higher priority to selecting peers except OSSs in ASs with OSSs. We evaluate the performance of the modified algorithms by simulation. Simulation results show that 1) MLH' and MPH' achieve almost the same maximum number of joining peers as MLH and MPH, respectively, 2) MLH' shows a maximum of 37% larger inter-AS traffic volume than MLH, 3) MPH' shows larger traffic volume than MPH when the number of joining peers is small, while MPH' shows a maximum of 38% smaller traffic volume than MPH when the number of joining peers is large thanks to the distribution of top level peers to many ASs.

Scopus

CiNii Article

researchmap

Language：Japanese  

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

CiNii Article

CiNii Books

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

CiNii Article

CiNii Books

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

CiNii Article

CiNii Books

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

With the spread of broadband access lines, many services making efficient use of them such as video streaming services have appeared. The popularity of these services can cause problems such as the shortage of the Internet backbone capacity, so it is important to detect increasing traffic which cause these problems early. In this paper, we propose an early detection method (partial aggregation method) of increasing traffic under per-subnet based distributed traffic monitoring. The method estimates future overall traffic volume by taking account of estimated traffic volume in all subnets and the estimated number of subnets generating increasing traffic. We compare the proposed method to a previous method (aggregation method) which takes account of estimated traffic volume in all subnets only. The results show that the partial aggregation method can detect increasing traffic earlier than the aggregation method by a maximum of 90 days.

CiNii Article

CiNii Books

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

With the spread of broadband access lines, many services making efficient use of them such as video streaming services have appeared. The popularity of these services can cause problems such as the shortage of the Internet backbone capacity, so it is important to detect increasing traffic which cause these problems early. In this paper, we propose an early detection method (partial aggregation method) of increasing traffic under per-subnet based distributed traffic monitoring. The method estimates future overall traffic volume by taking account of estimated traffic volume in all subnets and the estimated number of subnets generating increasing traffic. We compare the proposed method to a previous method (aggregation method) which takes account of estimated traffic volume in all subnets only. The results show that the partial aggregation method can detect increasing traffic earlier than the aggregation method by a maximum of 90 days.

CiNii Article

CiNii Books

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

As a new optical network architecture that can be built with currently available devices, light trail networks attract attention. A trail is an unidirectional communication channel which is set up by reserving a specific wavelength along a route between a source node and a destination node, and packet communication from an arbitrary node to its downstream nodes is possible. Because a trail is a shared medium, we need medium access control (MAC) to avoid collisions of packets. Token passing type MAC methods such as light-trail fair access MAC (LT-FA MAC) and adaptive round time MAC (ART MAC) have been proposed. In those methods, however, downstream bandwidth of the token holding node is only used, and upstream bandwidth becomes idle. In this paper, we try to improve throughput of a trail by using those idle bandwidth effectively in a token passing type MAC method. In order to achieve this, we propose a dynamic trail partitioning method that dynamically partitions a trail into upstream trail and downstream one at the token holding node, and permits parallel transfer of data using two trails. Numerical calculations and simulations show that the dynamic trail partitioning method can improve throughput by a maximum of more than 80%.

CiNii Article

CiNii Books

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

As a new optical network architecture that can be built with currently available devices, light trail networks attract attention. A trail is an unidirectional communication channel which is set up by reserving a specific wavelength along a route between a source node and a destination node, and packet communication from an arbitrary node to its downstream nodes is possible. Because a trail is a shared medium, we need medium access control (MAC) to avoid collisions of packets. Token passing type MAC methods such as light-trail fair access MAC (LT-FA MAC) and adaptive round time MAC (ART MAC) have been proposed. In those methods, however, downstream bandwidth of the token holding node is only used, and upstream bandwidth becomes idle. In this paper, we try to improve throughput of a trail by using those idle bandwidth effectively in a token passing type MAC method. In order to achieve this, we propose a dynamic trail partitioning method that dynamically partitions a trail into upstream trail and downstream one at the token holding node, and permits parallel transfer of data using two trails. Numerical calculations and simulations show that the dynamic trail partitioning method can improve throughput by a maximum of more than 80%.

CiNii Article

CiNii Books

J-GLOBAL

researchmap

researchmap

researchmap

Although the offset-time-based scheme can differentiate loss rates of packets according to their QoS levels in optical burst switching networks, it also increases end-to-end delay of packets with higher QoS level. This is because it increases offset time of bursts in proportional to QoS levels of packets which are included in the bursts. In this paper, we propose a burst assembly method that reduces end-to-end delay. Our proposed method includes packets that arrive during the offset time into the currently assembled bursts while the conventional method includes them into the burst assembled next time. Simulation results show that our method offers 12-23% delay reduction for packets with the highest QoS level.

Scopus

researchmap

Language：Japanese  

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

CiNii Article

CiNii Books

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

CiNii Article

CiNii Books

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

CiNii Article

CiNii Books

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

CiNii Article

CiNii Books

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

CiNii Article

CiNii Books

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

In this paper, we propose two peer selection algorithms to increase the number of joining peers and decrease the traffic amount between ASs in P2P live streaming, where we assume that the number (logical hop count) of relaying peers is limited by a predetermined value in order to keep real-time property of live streaming. In the first method, a newly joining peer selects such providing peers that the logical hop count is minimum to increase the number of joining peers. And then if there are several such providing peers, the newly joining peer selects such providing peers that the number (physical hop count) of ASs between the newly joining peer and each of the providing peers is minimum to decrease the traffic amount between ASs. In the second peer selection method, a newly joining peer selects providing peers in the reverse order of the first selection method. Simulation results show that the former increases the number of joining peers compared to the latter. Moreover, when the number of joining peers is small (large), the traffic amount between ASs in the latter (the former) is smaller.

CiNii Article

CiNii Books

researchmap

Language：English  

In this paper, we evaluate the performance of peer selection algorithms: MLH (Minimum Logical Hop) and MPH (Minimum Physical Hop) to increase the number of joining peers and to decrease inter-AS traffic volume in P2P live streaming, where we assume that every peer's logical hop count (the number of providing peers between an origin streaming server and the peer) is limited by a predetermined value in order to keep real-time property of live streaming. In MLH, a newly joining peer selects such providing peers that the logical hop count is minimum to increase the number of joining peers. And then if there are several such providing peers, the newly joining peer selects such providing peers that the number (physical hop count) of ASs between the newly joining peer and each of the providing peers is minimum to decrease the inter-AS traffic volume. In MPH, a newly joining peer selects providing peers in the reverse order of MLH. Simulation results show that MLH achieves about from 20% to 60% more maximum joining peers than MPH, and MPH shows smaller inter-AS traffic volume than MLH when the number of joining peers is small while MLH shows smaller inter-AS traffic volume when the number of joining peers is large. ©2009 IEEE.

DOI： 10.1109/APCC.2009.5375467

Scopus

researchmap

researchmap

CiNii Article

researchmap

Language：English   Publisher：ELECTRONICS TELECOMMUNICATIONS RESEARCH INST  

We evaluate performance of two medium access control methods with token passing, LT-FA MAC and ART MAC, in light trail networks. LT-FA MAC aims at coping with bursty packet arrivals while ART MAC aims at decreasing transmission delay. Simulation evaluation is performed in terms of the number of the maximum transmission nodes accommodated on a light trail and the maximum communication distance under the constraints that the bandwidth utilization rate of 90% and the average transmission delay of 100 ms must be achieved. Simulation results show that ART MAC permits about 500-1500km longer communication distance than LT-FA MAC with the same number of transmission nodes on a light trail and ART MAC accommodates about 15-25 more nodes than LT-FA MAC with the same communication distance.

Web of Science

CiNii Article

researchmap

Language：English  

In this paper, we evaluate the performance of peer selection algorithms: MLH (Minimum Logical Hop) and MPH (Minimum Physical Hop) to increase the number of joining peers and to decrease inter-AS traffic volume in P2P live streaming, where we assume that every peer's logical hop count (the number of providing peers between an origin streaming server and the peer) is limited by a predetermined value in order to keep real-time property of live streaming. In MLH, a newly joining peer selects such providing peers that the logical hop count is minimum to increase the number of joining peers. And then if there are several such providing peers, the newly joining peer selects such providing peers that the number (physical hop count) of ASs between the newly joining peer and each of the providing peers is minimum to decrease the inter-AS traffic volume. In MPH, a newly joining peer selects providing peers in the reverse order of MLH. Simulation results show that MLH achieves about from 20% to 60% more maximum joining peers than MPH, and MPH shows smaller inter-AS traffic volume than MLH when the number of joining peers is small while MLH shows smaller inter-AS traffic volume when the number of joining peers is large. ©2009 IEEE.

DOI： 10.1109/APCC.2009.5375467

Scopus

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

CiNii Article

CiNii Books

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

CiNii Article

CiNii Books

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

CiNii Article

CiNii Books

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

CiNii Article

CiNii Books

J-GLOBAL

researchmap

researchmap

Although Transmission Control Protocol (TCP) is widely used in the Internet, its performance is poor in networks with long delays. To improve TCP performance in such networks as long-delay satellite environments, the use of a PEP (Performance Enhancing Proxy) has been proposed. The PEP operates on a router that connects a terrestrial link and a satellite link along a TCP connection. When a data packet arrives at the PEP, it forwards the packet to the destination host, transmits the corresponding ACK (premature ACK) to the source host on behalf of the destination host, and stores a copy of the packet in a local buffer (PEP buffer) in case retransmission of the packet is required. As a congestion control method on the PEP, a method that keeps the number of prematurely acknowledged packets in the PEP buffer below a threshold (watermark) value has been proposed, and a watermark value optimization algorithm has been also proposed under the assumption that all connections have the same route. In this paper, we propose a watermark value optimization algorithm under a more realistic assumption that connections have different routes. We define the upper bound of the average throughput of a connection as the maximum average throughput in the case that the connection only exists in network. For a given set of connections, the proposed algorithm tries to attain that the average throughputs of connections with smaller upper bounds are equal to their respective upper bounds, and the number of such connections (maximized-connections) is maximized, and the average throughputs of connections with larger upper bounds are maximized under the condition that they are equal to each other and are larger than or equal to any upper bound of the maximized-connections. Numerical examples show that the proposed algorithm attains such goal. © 2007 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.

Scopus

researchmap

Language：Japanese  

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

In order to improve TCP performance in networks with large round trip time, a method using a PEP (Performance Enhancing Proxy) is a proposed. In the method, the PEP is located on a intermediate router along a TCP connection, and returns acknowlegdment packets (premture ACKs) to the source host instead of the destination host. In previous researches, although a congestion control in the PEP which keeps the number of prematurely acknowledged packets below a threshold (watermark) value has been proposed, packet losses are not taken into account. In this paper, we incorporate a retransmission control to deal with packet losses into the congestion control. the proposed retransmission control mainly has two functions. One is to return duplicate premature ACKs to solicit the fast retransmission in the source host, and the other is to retransmit packets by timeouts and receive of duplicate acknowledgments from the destination host. In addition, TCP selective acknowledgment (SACK) option is supported in the PEP. If the source host can treat the option, the PEP returns premature ACKs with the option.

CiNii Article

CiNii Books

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

In order to improve TCP performance in networks with large round trip time, a method using a PEP (Performance Enhancing Proxy) is a proposed. In the method, the PEP is located on a intermediate router along a TCP connection, and returns acknowlegdment packets (premture ACKs) to the source host instead of the destination host. In previous researches, although a congestion control in the PEP which keeps the number of prematurely acknowledged packets below a threshold (watermark) value has been proposed, packet losses are not taken into account. In this paper, we incorporate a retransmission control to deal with packet losses into the congestion control. the proposed retransmission control mainly has two functions. One is to return duplicate premature ACKs to solicit the fast retransmission in the source host, and the other is to retransmit packets by timeouts and receive of duplicate acknowledgments from the destination host. In addition, TCP selective acknowledgment (SACK) option is supported in the PEP. If the source host can treat the option, the PEP returns premature ACKs with the option.

CiNii Article

CiNii Books

researchmap

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

Change-point detection schemes are a promising approach for detecting network anomalies such as attacks and infections by unknown viruses and worms. They detect those behaviors as change-points. In general, however, because they also detect false-positive change-points, those caused by other factors such as hardware troubles, we need a scheme that only detects true-positive change-points caused by attacks and infections. True-positive change-points tend to occur simultaneously, and the number of true-positive change-points is very large, while false-positive change-points tend to occur sporadically. We exclude false-positive change-points by neglecting change-points that occur sporadically, based on information gathered from the whole network. In this paper, we propose a multi-stage network anomaly detection scheme that aggregates change-point information from distributed IDSs (Intrusion Detection Systems) and detects the true-positive change-points. Simulation results illustrate that, compared to a scheme using only one IDS, our method always yields a smaller false-positive rate, a reduction of up to 98%, under a constraint that the detection rate of the true-positive change-points must exceed 0.99.

CiNii Article

CiNii Books

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

Change-point detection schemes are a promising approach for detecting network anomalies such as attacks and infections by unknown viruses and worms. They detect those behaviors as change-points. In general, however, because they also detect false-positive change-points, those caused by other factors such as hardware troubles, we need a scheme that only detects true-positive change-points caused by attacks and infections. True-positive change-points tend to occur simultaneously, and the number of true-positive change-points is very large, while false-positive change-points tend to occur sporadically. We exclude false-positive change-points by neglecting change-points that occur sporadically, based on information gathered from the whole network. In this paper, we propose a multi-stage network anomaly detection scheme that aggregates change-point information from distributed IDSs (Intrusion Detection Systems) and detects the true-positive change-points. Simulation results illustrate that, compared to a scheme using only one IDS, our method always yields a smaller false-positive rate, a reduction of up to 98%, under a constraint that the detection rate of the true-positive change-points must exceed 0.99.

CiNii Article

CiNii Books

researchmap

Language：Japanese  

J-GLOBAL

researchmap

Language：Japanese  

J-GLOBAL

researchmap

Wavelength converters reduce the connection blocking probability in wavelength-routed networks by eliminating the wavelength continuity constraint. We develop a method for deployment of wavelength converters in wavelength-routed networks with an overlay model. In these networks, most wavelength converters are deployed on edge nodes to cover the difference in the numbers of wavelengths multiplexed on access and core links. Therefore reduction of wavelength converter cost on edge nodes leads to minimizing the wavelength converter cost in the whole network. We propose an ingress edge node architecture with fixed wavelength converters that have limited wavelength convertibility but are more economical than full wavelength converters. In our architecture, each input access link of ingress edge nodes is equipped with fixed wavelength converters, and input wavelengths from the access links are evenly distributed on the output core link. As a result, competition for a free wavelength on an output core link is avoided. Simulation results show that our edge node architecture offers about 20% cost reduction compared with a node architecture that uses only full wavelength converters where networks are actually under operation and a full wavelength converter cost to fixed wavelength converter cost ratio is 3:1. © 2006 Optical Society of America.

DOI： 10.1364/JON.5.000196

Scopus

researchmap

researchmap

Language：English   Publisher：IEEE  

To improve TCP throughput performance, a method using a PEP (Performance Enhancing Proxy) has been proposed. The PEP operates on a router along a TCP connection. When a data packet arrives at the PEP, it forwards the packet to the destination host, transmits the corresponding ACK (premature ACK) to the source host in behalf of the destination host, and stores a copy of the packet into its own buffer (PEP buffer) in case retransmission of the packet is required. As a congestion control method on the PEP, a method which keeps the number of prematurely acknowledged packets in the PEP buffer below a threshold (watermark) value has been proposed. However, the relation between the watermark value and throughput is not sufficiently investigated, and an optimization method of the watermark value is not proposed. In this paper, we first investigate the relation between the watermark value and the average throughput. Extensive simulations show that the simulation results are roughly classified into two cases. In the first case, the average throughput becomes larger for larger watermark values and becomes a constant value when the watermark is over a certain value. In the second case, although the average throughput becomes larger for larger watermark values in the same way, it decreases when the watermark is over a certain value. Next, based on the results about the relation, we propose an watermark optimization algorithm which can adaptively maximize the average throughput of each connection and also satisfy a fairness condition that the average throughputs of connections are equal to each other.

Web of Science

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

Wavelength converters reduce connection blocking probability in wavelength-routed networks by eliminating the wavelength continuity constraint. We develop a method for deployment of wavelength converters in wavelength-routed overlay networks. In these networks, most wavelength converters are deployed on edge nodes to cover the difference in the numbers of wavelengths multiplexed on access and core links. Therefore reduction of wavelength converter cost on edge nodes leads to minimizing the wavelength converter cost in the whole network. We propose an ingress edge node architecture with fixed wavelength converters that are more economical than full wavelength converters. In our architecture, each input access link of ingress edge nodes is equipped with fixed wavelength converters and input wavelengths from the access links are evenly distributed on the output core link. As a result, competition for a free wavelength on an output core link is avoided. Simulation results show that our edge node architecture offers more than 46% cost reduction.

CiNii Article

CiNii Books

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

Wavelength converters reduce connection blocking probability in wavelength-routed networks by eliminating the wavelength continuity constraint. We develop a method for deployment of wavelength converters in wavelength-routed overlay networks. In these networks, most wavelength converters are deployed on edge nodes to cover the difference in the numbers of wavelengths multiplexed on access and core links. Therefore reduction of wavelength converter cost on edge nodes leads to minimizing the wavelength converter cost in the whole network. We propose an ingress edge node architecture with fixed wavelength converters that are more economical than full wavelength converters. In our architecture, each input access link of ingress edge nodes is equipped with fixed wavelength converters and input wavelengths from the access links are evenly distributed on the output core link. As a result, competition for a free wavelength on an output core link is avoided. Simulation results show that our edge node architecture offers more than 46% cost reduction.

CiNii Article

researchmap

Language：English   Publisher：OPTICAL SOC AMER  

Many conventional WDM network planning and/or designing schemes assume that future traffic demand is known beforehand. However, it is difficult to predict future traffic demand accurately. We develop a scheme to design a WDM network that accommodates as much traffic as possible against a variety of traffic patterns, that is, a WDM network robust against traffic uncertainties. Our basic idea is to select a node pair that is expected to be a bottleneck in the future and then to deploy network equipment so that the volume of traffic accommodated by the node pair increases. Our simulation results show that the WDM network designed with our method can accommodate more traffic demand than those designed with the existing methods at the same cost. (c) 2004 Optical Society of America.

DOI： 10.1364/JON.4.000011

Web of Science

Scopus

researchmap

Language：English   Publisher：IEEE  

The scalability of routing protocol has been considered as a key issue in large-scaled wavelength routed networks. Hierarchical routing scales well by yielding enormous reductions in routing table length, but it also increases path length. This increased path length in wavelength-routed networks leads to increased blocking probability because longer paths tend to have less free wavelength channels. However, if the routes assigned to longer paths have greater wavelength resources, we can expect that the blocking probability will not increase. In this paper, we propose a distributed node-clustering method that maximizes the number of lightpaths between nodes. The key idea behind our method is to construct node-clusters that have much greater wavelength resources from the ingress border nodes to the egress border nodes. which increases the wavelength resources on the routes of lightpaths between nodes. We evaluate the blocking probability for lightpath requests and the maximum table length in simulation experiments. We find that the method we propose significantly reduces the table length, while the blocking probability is almost the same as that without clustering.

Web of Science

Scopus

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

The scalability of routing table length is indispensable for constructing large-scaled WDM network. Hierarchical routing yields enormous reduction in routing table length, but it also yields increase in path length. In wavelength-routed networks, the increase in path length leads to increase in blocking probability because of less probability to find idle wavelengths on the longer paths. However, if the longer paths have more wavelength resources, we expect that the blocking probability does not get worse. In this paper, we propose a new node-clustering method that maximizes the number of lightpaths between nodes. The key idea of our method is to construct the node-clusters that have much wavelength resources from the ingress border nodes to the egress border nodes, which increases the wavelength resources on routes of lightpaths between nodes. We evaluate blocking probability of lightpath requests and table length by simulation experiments. The results show that our proposed method greatly reduces the table length, while the blocking probability is almost the same as the results with no clustering.

CiNii Article

CiNii Books

researchmap

Language：Japanese  

J-GLOBAL

researchmap

Language：English   Publisher：The Institute of Electronics, Information and Communication Engineers  

Many researches have been investigated on planning or designing WDM networks assuming that the future traffic demand is known beforehand. Practically, it is difficult to predict the future traffic demand accurately since there are various types of data traffic with different traffic characteristics. In this paper, we propose a scheme to design a WDM network that will accommodate as much traffic as possible against a variety of traffic patterns, that is, robust WDM network. The straightforward way to meet this objective is to design a network to maximize the volume of traffic accommodated in the future. To achieve this simply, we divide the WDM network design problem into two subproblems, the OXC-deployment problem and the fiber-deployment problem. In both problems, we propose schemes to maximize the volume of traffic demand that can be accommodated in the future by taking into account the maximum flow value of each node-pair. We handle those problems by incrementally extending network resources based on the ADD algorithm. By deploying network resources until they reach a condition that a robust WDM network needs to fulfill, we can design a robust WDM network. We compare our proposed scheme with the existing method using various traffic matrices. The results show that the WDM network designed by our method can accommodate more traffic demand than the one, designed by the existing method with the same cost.

CiNii Article

CiNii Books

researchmap

Among the several architectures for IP (Internet Protocol) over WDM networks, one promising approach is to overlay a logical topology that consists of lightpaths on the physical WDM network, such that each lightpath carries IP traffic between edge nodes. Many of the conventional methods for designing the logical topology deal with the case of a constant number of wavelengths being available on each fiber. While the number of wavelengths offered by current optical technology is now of the order of thousands, it is not necessary to utilize all wavelengths on each fiber in building an effective logical topology on a WDM network. Instead,.several wavebands may be considered for introduction by deploying. optical fiber amplifiers when additional wavelengths are actually required. In this case, the number of available wavelengths on the respective fibers depends on the number of optical fiber amplifiers deployed on each fiber. In this paper, we propose a new heuristic algorithm for the design of a logical topology with as few optical fiber amplifiers as possible, so that the approach is cost-effective. We then compare our proposal with conventional methods in terms of average packet delays and throughput, and show that our algorithm reduces the number of optical fiber amplifiers required. We also investigate the effect of increasing the number of wavelengths multiplexed on each fiber, by assuming that the product of the number of available wavelengths and the bandwidth of each wavelength is constant. The results show that such a logical topology is able to accommodate a greater volume of traffic by multiplexing more wavelengths on each fiber. The approach thus conforms to Cao's Law. © 2003 by Springer Science+Business Media Dordrecht.

DOI： 10.1007/978-0-387-35670-9

Scopus

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

It is difficult to predict future traffic precisely although it is assumed that the amount of future traffic is predictable in many researches on a multi-period planning for WDM networks. In this paper, we propose a wave-length-routing algorithm that maximizes the number of accommodated lightpaths without precise prediction of future traffic demand. Since our algorithm reduces the number of leased fibers, it can be effectively applied to a multi-period planning for WDM networks. We show that the proposed algorithm reduces the number of leased fibers more than the existing algorithm does while it accommodates lightpaths more.

CiNii Article

CiNii Books

researchmap

© 2002 DTU. We propose a new routing algorithm for accommodating lightpaths in WDM networks. Our algorithm is integrated into multi-period planning of WDM networks without traffic prediction. We show that it achieves higher efficiency than existing algorithms.

Scopus

researchmap

Language：Japanese  

© 2002 DTU. We propose a new routing algorithm for accommodating lightpaths in WDM networks. Our algorithm is integrated into multi-period planning of WDM networks without traffic prediction. We show that it achieves higher efficiency than existing algorithms.

Scopus

J-GLOBAL

researchmap

Language：Japanese  

J-GLOBAL

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

Among several architectures for IP(Internet Protocol)over WDM networks, one promising way is to overlay a logical topology consisting of lightpaths over the physical WDM network such that each lightpath carries IP traffic between edge nodes. The conventional design methods of the logical topology consider the case that the constant number of wavelengths are available on each fiber. However, a current optical technology provides several wave-bands and the order of thousands wavelengths, by deploying additional optical fiber amplifiers, and hence the number of available wavelengths depends on the number of optical fiber amplifiers deployed on a fiber. In this paper, we propose a new heuristic algorithm to design a logical topology with a least number of optical fiber amplifiers. We then compare our proposal with conventional methods in terms of the average packet delays, throughput, and show that our algorithm reduces the number of optical fiber amplifiers. We also investigate the effect of increasing the number of wavelengths multiplexed on fiber, by asssuming the product of the number of available wavelengh and the bandwidth of a wavelength to be constant. The results show that the logical topology can accommodate more traffic volume by the large number of wavelengths multiplexed on the fiber, which confirms that the results yield a Cao's Law.

CiNii Article

CiNii Books

researchmap

Language：Japanese   Publisher：The Institute of Electronics, Information and Communication Engineers  

Among several architectures for IP(Internet Protocol)over WDM networks, one promising way is to overlay a logical topology consisting of lightpaths over the physical WDM network such that each lightpath carries IP traffic between edge nodes. The conventional design methods of the logical topology consider the case that the constant number of wavelengths are available on each fiber. However, a current optical technology provides several wave-bands and the order of thousands wavelengths, by deploying additional optical fiber amplifiers, and hence the number of available wavelengths depends on the number of optical fiber amplifiers deployed on a fiber. In this paper, we propose a new heuristic algorithm to design a logical topology with a least number of optical fiber amplifiers. We then compare our proposal with conventional methods in terms of the average packet delays, throughput, and show that our algorithm reduces the number of optical fiber amplifiers. We also investigate the effect of increasing the number of wavelengths multiplexed on fiber, by assuming the product of the number of available wavelength and the bandwidth of a wavelength to be constant. The results show that the logical topology can accommodate more traffic volume by the large number of wavelengths multiplexed on the fiber, which confirms that the results yield a Cao's Law.

CiNii Article

researchmap