In this paper, we analyze a performance model for the TCP Congestion Avoidance algorithm. The model predicts the bandwidth of a sustained TCP connection subjected to light to moderate packet losses, such as loss caused by network congestion. It assumes that TCP avoids retransmission timeouts and always has sufficient receiver window and sender data. The model predicts the Congestion Avoidance performance of nearly all TCP implementations under restricted conditions and of TCP with SelectiveAcknowledgements over a much wider range of Internet conditions. We verify

Although active networks have generated much debate in the research community, on the whole there has been little hard evidence to inform this debate. This paper aims to redress the situation by reporting what we have learned by designing, implementing and using the ANTS active network toolkit over the past two years. At this early stage, active networks remain an open research area. However, we believe that we have made substantial progress towards providing a more flexible network layer while at the same time addressing the performance and security concerns raised by the presence of mobile code in the network. In this paper, we argue our progress towards the original vision and the difficulties that we have not yet resolved in three areas that characterize a "pure" active network: the capsule model of programmability; the accessibility of that model to all users; and the applications that can be constructed in practice.

An analytical framework for multiple TCP flows sharing a bottleneck link under the Random Early Detection (RED) regime is developed. Closed form expressions for the steady state throughput and average queueing delay are derived and verified by simulations; these show that RED significantly improves the inherent TCP bias against links with higher round-trip delays as compared to Tail Drop, contrary to prevailing belief. Further, we derive closed form bounds on the minimum average queuing delay achievable through a RED gateway with no deterministic packet drop.

Abstract — The provision of quality of service (QoS) in a differentiated services (Diff-Serv) environment requires an adequate differentiation between high-priority/assured and low-priority/best-effort classes of service in the presence of congestion, giving priority/preference to assured-tagged traffic. For this purpose, a new active queue management scheme, implemented within the Diff-Serv framework, is presented that provides congestion control in TCP/IP networks using a fuzzy logic control approach. The proposed fuzzy logic approach for congestion control allows the use of linguistic knowledge to capture the dynamics of nonlinear probability marking functions, uses multiple inputs to capture the dynamic state of the network more accurately, and can offer effective implementation. A simulation study over a wide range of traffic conditions- considering multiple bottleneck links- shows that the fuzzy logic based controller outperforms the Random Early Detection (RED) implementation for Diff-Serv in terms of link utilization, packet losses, and queue fluctuations and delays. Also, the proposed scheme can offer better differentiation among assured and best-effort traffic, thus it can provide better QoS to different types of data streams, such as TCP/FTP traffic or TCP/Web-like traffic, whilst maintaining high utilization. I.

We investigate the service differentiation, in terms of average throughput, and the performance achieved using weighted window-based congestion control in networks supporting Explicit Congestion Notification (ECN). Our results show how service differentiation, queueing delay, and average throughput are affected by the increase and decrease rules of the end-system congestion control algorithms, and how they depend on the marking algorithms operating in the routers. The end-system algorithms we investigate include WTP (Willingness-To-Pay) and MulTCP congestion control, and the packet marking algorithms include RED, virtual queue marking, and load-based marking. Our investigations consider both single and multiple link topologies, and connections with different round trip times.

Several objectives have been identified in developing the randomearly drop (RED): decreasing queueingdelay increasing throughput, and increasing fairness between short and long lived connections. It has been believed that indeed the dropprobability of a packet in RED does not depend on the size of the file to which it belongs. In this paper westudy the fairness properties of RED where fairness is taken with respect to the size of the transferred file. We focus on short lived TCP sessions. Our findings are that (i) in terms of loss probabilities, RED is unfair: it favors short sessions, (ii) RED is fairer in terms of the average throughput of a session (as a function of its size) than in terms of loss probabilities.

TCP is known to have poor performance over unreliable wireless links where packet losses due to transmission errors are misinterpreted as indications of network congestion. TCP enhancements proposed in the literature differ in their signaling and data recovery mechanisms, applicable network configurations, traffic scenarios and locations where required changes are made. In this paper we categorize existing enhancements into several approaches. Motivated by these criteria, we propose a new enhancement that requires only local changes, but applies to a broad range of network and traffic configurations. Comparison with existing algorithms show this new enhancement achieves excellent performance.

Abstract: Most of the studies on TCP assume a fixed number of persistent TCP connections. A few of the recent studies look at non-persistent TCP connections but with many limiting assumptions. We look at a bottleneck router which has real time UDP traffic, TCP ON-OFF traffic and a stochastic flow of non-persistent TCP connections going through it. Our analysis is able to handle a large number (100s) of TCP connections and still provides the mean file download times of each of these connections when all of them have different packet sizes, round trip times and max window size. Using this approach we are able to model almost all the different traffic types that exist in the Internet today. We also study systems employing RED control, which is recommeded to be used in the Internet. We demostrate the effectiveness of RED for congestion control in realistic Internet scenarios both theoretically and via simulations. Keywords: Persistent and non-persistent TCP connections, performance analysis, mean download time, Internet, RED. 1

Today's known and widely used active queue management (AQM) schemes do not differentiate between packets from responsive (e.g., TCP sessions) and non-responsive traffic (e.g., UDP). This results in further widening the gap of unfair advantage already inherent to non-responsive traffic, as the responsive sender will significantly reduce its future transmit rate as a result of the congestion signals. As a simple work-around, responsive and non-responsive traffic are often assigned distinct AQM parameters. This approach however requires tuning for each traffic class that potentially depends on the current or expected offered load. In other words, responsiveness and TCPfriendliness cannot be estimated easily --- not at last due to short-lived TCP sessions. In this paper we propose a closed-loop congestion control (CLCC) scheme on top of an existing AQM scheme to achieve fair bandwidth distribution among concurrent responsive and non-responsive traffic. The new scheme has the advantage that it does not need to estimate the level of responsiveness of traffic. We analyze our scheme on top of an existing rate-based AQM scheme known to approximate max-min fairness, and by means of simulations show that our extension significantly improves fair bandwidth allocation for responsive and non-responsive traffic. The simulation results have been verified with a prototype implementation on the IBM PowerNP 4GS3 network processor.

Explicit Congestion Notification (ECN) for IP networks has received considerable attention in recent years, and has been shown to improve TCP goodput. Previous studies have centered on scenarios in which TCP with ECN (TCP/ECN) traffic competes with ECN-unaware traffic. This paper presents case studies in which moderately short flows are all ECN-capable, and compare them with the corresponding cases where the flows are ECNunaware, running over drop tail and Random Early Detection routers. Using transmission overhead metrics, we show that TCP/ECN uses network resources more efficiently. We also consider the case of battery-operated devices and show that TCP/ECN is more power conserving than standard TCP. An unexpected outcome of our experiments is that goodput does not improve in an all-TCP/ECN environment.