A video streaming user is usually interested in some semantically-important scenes of a video content, e.g. goal scene in a football game. If the user could watch the video segments consisting that scene with high quality, the quality of experience (QoE) of the user would be substantially improved more than ever. Although Adaptive Bitrate (ABR) algorithm considering this aspect can achieve better QoE, design of current approach has been limited only inside the application layer. It is expected that higher QoE could be achieved when the underlying transport layer also handle these semantics information for its congestion control mechanism. In this paper, we propose a new concept for quality of service (QoS) control with differentiated explicit congestion notification (ECN) to improve QoE and reveal that throughput differentiation by ECN considering video semantics can improve QoE of a video streaming user. Evaluation results show that throughput differentiation by congestion control conducted inside the network improves the QoE of the user up to 57.9%.

In multicast RDMA, the sender overestimates congestion signals when multiple paths are congested. This congestion path multiplicity problem is hard to be resolved because RDMA sender cannot identify the sender of congestion signals due to RDMA protocol regulation. In this paper, we propose a new approach that the branching switch filters congestion signals from the worst congested interface. Our evaluation results show that our proposed method can adequately resolve the congestion path multiplicity problem.

In a data center (DC) where inter- and intra-DC flows share high-speed data center networks (DCNs), these two types of flows have significantly different round-trip times (RTTs). When congestion occurs in a DCN, the congestion signal is detected first by the intra-DC flow, resulting in unfairness where only it reacts to congestion. This technical problem is especially important for the receiving DCN because the congestion point is far from the sender. To solve this issue, we propose DC performance enhancing proxy (DC-PEP) with ECN (Explicit Congestion Notification) buffer control as a proxy at the gateway of the receiving DCN. DC-PEP divides the end-to-end congestion control loop into two loops, which homogenizes the congestion loop inside the receiving DCN. To absorb the throughput difference between these two congestion loops, a buffer is required at DC-PEP. By implementing ECN-Echo notification, excessive queueing at DC-PEP can be avoided. Our simulation results show that our proposed DC-PEP with ECN improves fairness between inter- and intra-DC flows without incurring large queue sizes at the proxy.

A practical congestion control scheme for Named Data Networking (NDN), called PCON, has been proposed for efficient data retrieval from multiple sources. Although PCON has been proposed for a multi-path/-source data retrieval, it does not discuss resource pooling viewpoint which is an important fairness concept in resource sharing. In this paper, we first evaluate the fairness performance of PCON and show that PCON cannot achieve the global fairness based on the resource pooling concept. To improve the fairness, we proposed a method that enhances AQM mechanisms of PCON to improve fairness among flows sharing multiple communication resources. Through performance evaluation using a ns-3-based NDN simulator, we reveal that our proposed method can balance the congestion, improve the global fairness, and achieve high-level performance from the resource pooling perspective.

BBR has been proposed as a new congestion control algorithm avoiding bufferbloat and is one of the widely used one along with CUBIC in the current Internet. In this paper, we evaluate preliminary performance of BBR when it shares the bottleneck link with CUBIC. Evaluation results reveal that one of the reasons for unfairness between BBR and CUBIC is misestimation of BBR control parameter RTprop. We also reveal that by increase/decrease of RTprop, we can control delivery rate of BBR. Based on these insights, we propose a new BBR algorithm, BAR(BBR with Adjusting RTprop). BAR adjusts RTprop so that BBR obtains fair bandwidth with CUBIC. Our performance evaluation results show that inter-protocol fairness between BBR and CUBIC flows can be significantly improved by BAR with few packet losses.

User perceived performance of MPEG DASH video streaming deeply depends on interaction of higher layer (application layer) behavior of ABR algorithm and lower layer (transport layer) behavior of TCP algorithm. The lower layer behavior of MPEG DASH is determined by a TCP version, e.g. CUBIC and Compound TCP. The higher layer behavior of MPEG DASH depends on an ABR algorithm for bitrate selection. In this work-in-progress paper, we would like to show a potential approach of fairness improvement for carefully selected combination of the ABR algorithm and the TCP version, which is obtained from our on-going evaluation results.

In recent years, video traffic has rapidly increased, and reducing video traffic is an important issue for network providers. By caching video content at cache servers close to users, network providers can expect to reduce the video traffic in the networks. However, the storage capacity of cache servers is limited, so it is necessary to carefully select contents to be cached to effectively utilize the limited cache resources. In order to make effective use of cache resources, it is important to cache content based on the popularity dynamics of video contents. It is known that video contents have different popularity dynamics in each video category. For example, videos of movie and music categories tend to maintain view counts over long time, whereas the view counts of videos of news and sports categories rapidly decrease. In this paper, we propose a caching method that selects video content to be cached based on the popularity dynamics of video content in each category. To clarify the effectiveness of the proposed caching method, we evaluate the cache hit ratio of the proposed method by a trace-driven simulator using a measured request pattern of YouTube videos. We show that the proposed method improves the cache hit ratio compared with the LRU.

In recent years, the majority of internet traffic is video traffic such as YouTube and NetFlix, and reducing video traffic is an essential subject for network providers. Placing copy of original video contents in replication servers close to users has a chance to reduce the traffic in the networks. However, the storage capacity of a replication server is generally limited, so it is necessary to select video contents which will be requested in the near future. In the existing researches, temporarily or regionally popular contents are selected by focusing on the past popularity. In this paper, we would like to focus on another aspect of content request characteristics, user taste, and we propose to place contents which are not so popular but expected to be viewed by other users having similar tastes in the near future. Users with similar taste tend to request similar video contents, so we propose to select contents based on the measured tastes of users to place in the storage of replication servers. In the proposed method, cache storage is divided into two parts, a large static part and a small dynamic part. Cache-decision policy using user tastes is applied for the static part. In the dynamic part, LRU (Least Recently Used) cache-decision policy is applied to store contents having high popularity just after they are generated. The numerical results using MovieLens dataset show that the proposed method can make user groups with similar taste and improve the cache hit ratio compared with the LRU.

In Adaptive Bit Rate (ABR) utilized in MPEG-DASH (MPEG Dynamic Adaptive Streaming over HTTP), there have been proposed three representative algorithms, rate-based, buffer-based and hybrid-based ones. In the information centric networks, which have been widely studied as a network architecture for efficient content distribution including video streaming, routers are generally equipped with cache storage. When some segments of a required content is occasionally downloaded from cache storage in a router, content download from a router might be faster than segment retrieval from the original content server. This inaccurate throughput estimation might lead to inappropriate bitrate selection of ABR algorithm running on the user's application and significant degradation of QoE (Quality of Experience) performance. In this paper, from the viewpoint of QoE, we evaluate the influence of in-network caches on three types of ABR algorithm in video streaming.

Resource Pooling concept has been proposed as fair sharing of total network resources among all users sharing a network so that a whole network is treated as a single pooled resource. MPTCP (Multipath TCP) has been proposed as one of the most promising congestion controls which realize the Resource Pooling in the situation that each user can utilize multiple paths. In CCN (Content-Centric Networking), several multipath/multisource congestion control methods have been proposed. A fundamental difference between CCN congestion control and MPTCP where single source is assumed, is multisource and the Resource Pooling in multipath/multisource situation is quite a new research issue. In this paper, we discuss congestion control for CCN from the viewpoint of the Resource Pooling and show that our previously proposed congestion control satisfies the Resource Pooling concept. Our simulation results also show that our proposed congestion control achieves macroscopic fairness among users, which means the Resource Pooling concept is satisfied.

ICN (Information-Centric Networking) originally innovated for efficient content distribution, is now discussed to be applied to edge computing in IoT (Internet of Things) environment. In this paper, we focus on more flexible network processing environment, in-network processing, which is realized with ICN architecture. In our assumed environment, multiple functions are executed on different routers widely distributed in a whole network and the end-to-end optimal route for any data processing should be selected to satisfy various IoT applications' requirement. Our proposal, an on-demand routing method efficiently chains data and multiple functions compared to an existing proactive routing method. Also, our method reactively caches routing information in the network and realize scalable routing for ICN-based in-network processing.

This paper proposes a new cache management method considering the popularity of contents for in-network caching. In networks with in-network caching, routers store contents in their cache and users can download them from the routers. By utilizing cached contents, the loads for content servers and networks can be reduced. The proposed method provides a distributed cache management strategy that aims at utilizing unpopular contents in addition to popular contents. Through simulation experiments, we show that the proposed method enhances the cache utilization efficiently.

Multicast communications in data center has gotten great attention with its requirement for social networking services and file distribution, such as OS update towards virtual machines. Multicast communications offer advantage of efficient network resource usage. To the best of authors' knowledge, only feasibility and scalability issues have been treated in published papers for multicast communications in data centers. Data center networks generally operate at high utilization, so congestion control is one of the most important technical issues. This is also the case for multicast communications, but there have not been published any papers related to multicast congestion control in data centers. In this paper, we propose a new multicast congestion control, DCMC (Data Center Multicast Congestion control). DCMC is an end-to-end multicast congestion control and adjusts the sending rate to the estimated throughput of the worst congested receiver. The worst congested receiver is selected by equation-based estimation of receiver's throughput. Our simulation results show that DCMC can achieve TCP friendliness, i.e. fairness, with DCTCP with keeping high link utilization.

Network traffic for sharing contents is significantly increasing. The cache network is a desirable architecture to reduce traffic and to improve reliability of contents retrieval. “Breadcrumbs” is one of the promising techniques to find cached contents in a distributed manner. Thanks to Breadcrumbs, users can retrieve a content without managing where the cache exists. This demo shows a contents sharing application among mobile users in the Breadcrumbs-enabled cache network.

In wireless multi-hop networks, content servers are generally located outside a wireless multi-hop network and a user accesses these servers through a gateway node. So, content traffic has a tendency to be concentrated at a gateway node, which might cause throughput degradation in a whole wireless multi-hop network. Caching network is one promising way to resolve this technical problem. However, caching network has a limitation of cache availability on a default-path. For wired networks, Breadcrumbs has been proposed as in-network guide for cached contents. In this paper, we evaluate Breadcrumbs in wireless multi-hop networks. Our simulation results show that Breadcrumbs improves total throughput performance of wireless multi-hop networks. And also they surprisingly show that not only popular content throughput but also unpopular content throughput is improved.

Content transfer, e.g. video transfer, is now becoming major traffic in the current Internet. Content transfer originally has a content oriented feature of “wherever content is retrieved, users only take care of what content they obtain”. Conventional traffic engineering aims to obtain optimal routes for traffic between ingress and egress router pairs, which has “location oriented feature, i.e. where to connect”. With content oriented feature of user demand for content traffic, traffic engineering can be changed towards content oriented routing design. In this paper, we propose a novel content oriented traffic engineering, named CAR (Content Aware Routing). In CAR, routes are designed for content and egress router pair, i.e. content traffic towards a receiver-side router. Content demand can be flexibly distributed to multiple servers(repositories) providing the same content, which means content can be obtained from anywhere. With numerical examples by linear programming formulation, we comparatively evaluate CAR with comparing with combination of CDN and traffic engineering, i.e. location oriented designs. Numerical results show that CAR improves maximum link utilization significantly with slight increase of network resource usage.

In recent years, much work has been devoted to developing protocols and architectures for supporting the growing trend of data-oriented services. One drawback of many of these proposals is the need to upgrade or replace all the routers in order for the new systems to work. Among the few systems that allow for gradual deployment is the recently-proposed Breadcrumbs technique for distributed coordination among caches in a cache network. Breadcrumbs uses information collected locally at each cache during past downloads to support in-network guiding of current requests to desired content. Specifically, during content download a series of short-term pointers, called breadcrumbs, is set up along the download path. Future requests for this content are initially routed towards the server which holds (a copy of) this content. However, if this route leads the request to a Breadcrumbs-supporting router, this router re-directs the request in the direction of the latest downloaded, using the aforementioned pointers. Thus, content requests are initially forwarded by a location ID (e.g., IP address), but encountering a breadcrumb entry can cause a shift over to content-based routing. This property enables the Breadcrumbs system to be deployed gradually, since it only enhances the existing location-based routing mechanism (i.e. IP-based routing). In this paper we evaluate the performance of a network where Breadcrumbs is only partially deployed. Our simulation results show Breadcrumbs performs poorly when sparsely deployed. However, if an overlay of Breadcrumbs-supporting routers is setup, system performance is greatly improved. We believe that the reduced load on servers achieved with even a limited deployment of Breadcrumbs-supporting routers, combined with the flexibility of being able to deploy the system gradually, should motivate further investigation and eventual deployment of Breadcrumbs.

In wireless multi-hop networks, when an end host would like to communicate with external wired hosts, data packets are to be transmitted at first to a gateway node which is connected to both a wireless multi-hop network and a wired network. When there are multiple candidate gateways to be connected, a wireless host should select one gateway as its own default gateway. One possible and simple way is minimum hop gateway selection policy. This minimum hop gateway selection may cause performance degradation in the case of imbalance of traffic distribution. In this paper, we propose a new gateway selection method where an end host selects the default gateway from the viewpoint of gateway load balancing. Wireless communication channel around the gateway is very important resource because all the external traffic goes through the gateway. So, gateway load in our proposed method takes account of not only traffic intensity at a gateway but also interference of wireless channel around a gateway. Our simulation results show that our proposed method brings improvement of gateway throughput performance.

In ad hoc networks, a TCP session can obtain very low throughput with increase of the number of hops. One promising way to resolve this technical problem of TCP throughput degradation is proxy-based TCP. In proxy-based TCP, an end- to-end TCP session is divided by (an) intermediate proxy host(s), which leads to reduction of RTT of each divided session and improvement of TCP throughput. We evaluate proxy-based TCP in ad hoc networks with a realistic assumption of finite buffer in proxy hosts. Our simulation results show that proxy-based TCP surprisingly does not bring performance improvement and obtain lower throughput than conventional end-to-end TCP. In the paper, we show that wireless channel capture around the proxy host is the reason for this degradation. And we propose adaptive rate control method and intentional flow control method at the proxy host which improves throughput performance of proxy- based TCP. Performance evaluation results show that proxy- based TCP with our proposed methods outperforms conventional end-to-end TCP.

Application-level multicast (ALM) is a feasible alternative to IP multicast. In ALM, multicast related features, such as group membership management, multicast routing and packet replication, are implemented at end-hosts instead of routers. A multicast distribution tree is constructed in the application layer, so all nodes in this tree are end-hosts. Packet transmission between end-hosts uses conventional IP unicast service. Therefore, all end-hosts can enjoy multicast communications without IP multicast service. However, ALM has a serious problem that the multicast distribution tree is intrinsically fragile and an end-host failure causes tree partitions. In this paper, to deal with this problem, we propose a new tree construction protocol which makes outdegrees of intermediate nodes be balanced. The degree-balanced distribution tree can reduce the average number of nodes decoupled by tree partitions. To investigate performance of our protocol, it is compared with an existing ALM protocol. Our simulation results show that our protocol outperforms the existing protocol from the viewpoints of robustness, loss probability and receiver-perceived delay.

In this paper, we discuss TCP congestion control in the future heterogeneous mobile Internet where various access links such as a wireless LAN and a cellular network are integrated. Vertical handover (VH), which is a handover technique between different types of access links, is one of the most important technologies. However, the vertical handover significantly degrades the performance of TCP because it causes a drastic change of the bandwidth-delay product (BDP). In this paper, we propose a new congestion control for improving TCP performance during vertical handover. The proposed system is receiver driven and the TCP receiver controls the sending rate by ACK pacing. With computer simulation, we have shown that the proposed system adjust its congestion window size to a target BDP value rapidly and can achieve a good throughput performance.

TCP traffic has been reported to have not only self-similarity but also a multi-fractal nature. The multi-fractal nature of traffic volume in general is well known to have a significant influence on queueing behavior in short time scales. In our previous study, we showed that the retransmission time out (RTO) mechanism plays an important role in the multi-fractal nature of TCP traffic. In this paper, we investigate how to control the multifractal nature of TCP traffic in a real network. Here we focus on the random early detection (RED) mechanism in order to reduce the occurrence of RTO. We evaluate the RED mechanism from the viewpoint of its multi-fractal nature and show that the RED mechanism improves not only throughput and delay performance, but also the multi-fractal nature of TCP traffic.

TCP traffic has been reported to have not only self-similarity but a multi-fractal nature. The multi-fractal nature of a traffic volume, in general, is well known to have a significant influence on queueing behavior in the short time scale. We investigate what component of the TCP congestion control mechanism affects the multi-fractal nature of generated traffic. We artificially divide the TCP congestion control mechanism into several components and evaluate the multi-fractal nature of the generated traffic based on multi-fractal scaling analysis. Our simulation results show that the retransmission time out mechanism plays an important role in the multi-fractal nature of TCP traffic.

In content distribution networks(CDNs), the content routing which directs user requests to an adequate server from the viewpoint of improvement of latency for obtaining contents is one of the most important technical issues. Several information, e.g. server load or network delay, can be used for content routing. Network support, e.g. active network, enables a router to select an adequate server by using these information. We investigate a server selection policy of a network support approach from the viewpoint of which information to be used for effective server selection. We propose a server selection policy using RTT information measured at a router. Simulation results show that our proposed server selection policy in content routing selects a good server under both conditions where server latency and network delay is a dominant element of user response time. Furthermore, we also investigate the location of routers with network support bringing good performance for our proposed scheme.

A significant amount of wireless traffic is expected to be carried in the Internet, and wireless connections need to share the network resources with wired connections. However, in a wireless network environment, TCP, one of the most important transport protocols of TCP/IP, suffers from significant throughput degradation due to the lossy characteristics of a wireless link. Therefore, in order to design the next generation mobile networks, it is necessary to know how much the wireless connection suffers from the degradation in comparison to the wired connection. We discuss the fairness issue between TCP connections over wireless and wired links, and theoretically analyze throughput fairness between TCP over a wireless link with ARQ (automatic repeat request) based link layer error recovery and TCP over an error-free wired link. We validate our analysis by comparing numerical results obtained from the analysis with computer simulation ones.

One of the most important technical problems in reliable multicast protocols is reducing redundant NAKs (negative acknowledgements) to avoid NAK implosion. A number of NAK suppression mechanisms have been proposed to deal with this problem. In MBONE, which is a virtual multicast network and makes multi-point communication across the Internet feasible, the source link, the links directly connected to or very close to the source, contributes 5% packet loss. In the NAK suppression mechanism, in the case of such a loss, the source link loss, all receivers suffer the same packet loss and the NAK suppression mechanism does not work effectively. In this paper we propose a reliable multicast protocol applied local FEC, called the local FEC, where the source link loss is recovered with the FEC applied locally only to the source link. To investigate the performance of the local FEC, it is compared with the reliable multicast protocol with a conventional NAK suppression mechanism. Our simulation results show that the local FEC outperforms the NAK suppression protocol from the view point of scalability and wasted network bandwidth.

Previous measurement showed that self-similar nature is found not only in network traffic volume but also round trip packet delay. In this paper, we discuss three issues of the self-similarity of round trip time (RTT), which is one of the most important parameters to determine TCP throughput performance. First, we discuss the origin of the packet delay self-similarity. One study anticipated that the queueing delay of self-similar traffic is the reason for packet delay self-similarity. With computer simulation, we evaluate the correlation between traffic and RTT self-similarity. Next, we investigate the impact of RTT self-similarity on TCP throughput performance. Computer simulation results show that RTT self-similarity gives high variability to file transfer time. Finally, we investigate the impact of RTT self-similarity on RTO (retransmission time out). We discover that the bigger the Hurst parameter of the RTT is, the more frequent unnecessary timeouts occurs. Furthermore, we propose a new RTO calculation algorithm to improve these unnecessary timeouts.

One of the most important technical problems in reliable multicast protocol is to reduce redundant control packets and retransmitted packets. An approach for reducing them by making use of newly-invented network architecture, the active network technology, has been proposed. In the paper we compare a network-initiated approach, i.e. a reliable multicast protocol with network support, and conventional end-to-end approaches. These are compared under more realistic situations such as part of routers are equipped with the newly-invented active network technology. Our simulation results show that the network-initiated approach with 15% active routers outperforms conventional end-to-end approaches from the viewpoint of scalability. We also investigate the desirable location of active routers with a sophisticated network topology model, Tiers model, which reflects hierarchical structure of the Internet. Our simulation results suggest that at first active routers should be implemented in WANs because delay performance can be improved with a small number of active routers. After active routers are broadly implemented in WAN, the policy for location of active routers should be switched to MAN with expecting a decrease of redundant NAK transmission inside a network.

A growing number of network applications require the use of a reliable multicast protocol to disseminate data from a source to a potentially large number of receivers. This paper presents an analytic performance analysis of the packet delay incurred under three generic sender- and receiver-initiated approaches towards reliable multicast. We focus on the host processing requirements of these protocols and derive expressions for average time between the initial arrival of a packet at a sender and its correct reception at a randomly chosen receiver. Our numerical results indicate that a NAK-based protocol that limits NAK generation by intentionally and randomly delaying NAK packets can achieve substantially higher throughput than the other two protocols examined and can do so without suffering an appreciable higher delay over a range of system parameters.

Using the rate control to restrict the peak cell rate of traffic is addressed. The asynchronous transfer mode (ATM) environment is considered as a discrete-time environment in which one slot is equal to one cell transmission time. A rate control that can restrict the number of cells during any fixed-length time interval in a discrete-time environment is studied. The scheme is realized by means of simple counter accumulation per slot, and it is easy to synchronize with slots. The simplicity of this scheme may make possible the flexible restriction of the peak cell rate according to network conditions. The scheme is analyzed, using matrix analytic techniques, in order to investigate the smoothing effect of bursty traffic and the queue length required to satisfy a certain quality of cell loss probability. Numerical results show the effect of restricting the peak cell rate of bursty traffic and the tradeoff between the smoothing effect and the required queue length.>

A traffic control scheme for FDDI (fiber distributed data interface) interconnection through an ATM (asynchronous transfer mode) network is presented. The scheme consists of two parts, a VP (virtual path) bandwidth allocation algorithm and a feedback type buffer protection flow control. Both use the acknowledge for the frame of FDDI as a control unit. Simulation results show that VP bandwidth allocated by the scheme increases according to arrival traffic when the destination FDDI has enough capacity and allocated bandwidth is not increased when its capacity is less than arrival traffic. Thus, the scheme has desirable VP bandwidth allocation characteristics. Simulation results also show that the scheme prevents cell loss at a destination gateway completely.>

The authors propose two multicast routing algorithms which distribute copy operation of packets over all nodes along the multicast path: a link-added type algorithm and a loop-constructed type algorithm. Both algorithms, at first, derive an approximate solution for minimum cast path, and then improve the solution to prevent concentration of packet copy operation at one switching node at a little sacrifice of total cost along the path. Computer simulation results show that too much copy operation per node can be avoided by these algorithms. Compared to the minimum cost solution of the tree-shaped multicast path, the solution of the proposed algorithms makes the average distance connecting a source-destination pair longer, but the sacrifice of total cost is verb small. These algorithms can be applied not only to packet networks but also asynchronous transfer mode (ATM) networks.>

An asynchronous transfer mode (ATM) switch architecture that uses the broadcasting transmission medium for transmission of cells from input ports to output ports is introduced. Cell transmission and its control are separated completely, and cell transmission control, i.e. header operation, is executed before cell transmission (control ahead). With this operation, cell transmission and its control can be executed in a pipeline style, allowing high-speed cell exchange and making transmission control easier. One of the essential problems for ATM switches which use the broadcasting transmission medium is high-speed operation of the transmission medium. The switch fabric performance is analyzed according to its switching speed. Numerical results show that the ATM switch proposed shows good cell loss performance even when its switching speed is restricted, provided that switch utilization is below 1. Extensions to the switch that lead to robustness against bursty traffic are shown.>