Transport Control Protocol Throughput Research Articles

Implementation of the Algorithm for Congestion control in the Dynamic Circuit Network (DCN)

Transport Control Protocol (TCP) incast congestion happens when a number of senders work in parallel with the same server where the high bandwidth and low latency network problem occurs. For many data center network applications such as a search engine, heavy traffic is present on such a server. Incast congestion degrades the entire performance as packets are lost at a server side due to buffer overflow, and as a result, the response time becomes longer. In this work, we focus on TCP throughput, round-trip time (RTT), receive window and retransmission. Our method is based on the proactive adjust of the TCP receive window before the packet loss occurs. We aim to avoid the wastage of the bandwidth by adjusting its size as per the number of packets. To avoid the packet loss, the ICTCP algorithm has been implemented in the data center network (ToR).

ICTCP: Incast Congestion Control for TCP in Data-Center Networks

Transport Control Protocol (TCP) incast congestion happens in high-bandwidth and low-latency networks when multiple synchronized servers send data to the same receiver in parallel. For many important data-center applications such as MapReduce and Search, this many-to-one traffic pattern is common. Hence TCP incast congestion may severely degrade their performances, e.g., by increasing response time. In this paper, we study TCP incast in detail by focusing on the relationships between TCP throughput, round-trip time (RTT), and receive window. Unlike previous approaches, which mitigate the impact of TCP incast congestion by using a fine-grained timeout value, our idea is to design an Incast congestion Control for TCP (ICTCP) scheme on the receiver side. In particular, our method adjusts the TCP receive window proactively before packet loss occurs. The implementation and experiments in our testbed demonstrate that we achieve almost zero timeouts and high goodput for TCP incast.

TCP and web browsing performance in case of bi-directional packet loss

Performance modeling of the transport control protocol (TCP) has received a lot of attention over the past few years. The most commonly quoted results are approximate formulas for TCP throughput (Padhye et al. (2000) [1]) and document download times (Cardwell et al. (2000) [2]) which are used for dimensioning of IP networks. However, the existing modeling approaches unanimously assume that packet loss only occurs for packets from the server to the client, whereas in reality the packets in the direction from the client to the server may also be dropped. Our simulations with NS-2 show that this bi-directional packet loss indeed may have a strong impact on TCP performance. Motivated by this, we refine the models in Padhye et al. (2000) [1] and Cardwell et al. (2000) [2] by including bi-directional packet loss, also including correlations between packet loss occurrences. Simulations show that the proposed model leads to strong improvements of the accuracy of the TCP performance predictions. In addition we show how our model can be used to predict quality of experience for web browsing sessions.

Analytical modeling of TCP dynamics in infrastructure-based IEEE 802.11 WLANs

IEEE 802.11 wireless local area network (WLAN) has become the prevailing solution for wireless Internet access while transport control protocol (TCP) is the dominant transport-layer protocol in the Internet. It is known that, in an infrastructure-based WLAN with multiple stations carrying long-lived TCP flows, the number of TCP stations that are actively contending to access the wireless channel remains very small. Hence, the aggregate TCP throughput is basically independent of the total number of TCP stations. This phenomenon is due to the closed-loop nature of TCP flow control and the bottleneck downlink (i.e., access point-to-station) transmissions in infrastructure-based WLANs. In this paper, we develop a comprehensive analytical model to study TCP dynamics in infrastructure-based 802.11 WLANs. We calculate the average number of active TCP stations and the aggregate TCP throughput using our model for given total number of TCP stations and the maximum TCP receive window size. We find out that the default minimum contention window sizes specified in the standards (i.e., 31 and 15 for 802.11b and 802.11a, respectively) are not optimal in terms of TCP throughput maximization. Via ns-2 simulation, we verify the correctness of our analytical model and study the effects of some of the simplifying assumptions employed in the model. Simulation results show that our model is reasonably accurate, particularly when the wireline delay is small and/or the packet loss rate is low.

Robust end-to-end loss differentiation scheme for transport control protocol over wired/wireless networks

The authors propose a robust end-to-end loss differentiation scheme to identify the packet losses because of congestion for transport control protocol (TCP) connections over wired/wireless networks. The authors use the measured round trip time (RTT) values to determine whether the cause of packet loss is because of the congestion over wired path or regular bit errors over wireless paths. The classification should be as accurate as possible to achieve high throughput and maximum fairness for the TCP connections sharing the wired/wireless paths. The accuracies of previous schemes in the literature depends on varying network parameters such as RTT, buffer size, amount of cross traffic, wireless loss rate and congestion loss rate. The proposed scheme is robust in that the accuracy remains rather stable under varying network parameters. The basic idea behind the scheme is to set the threshold for the classification to be a function of the minimum RTT and the current sample RTT, so that it may automatically adapt itself to the current congestion level. When the congestion level of the path is estimated to be low, the threshold for a packet loss to be classified as a congestion loss is increased. This avoids unnecessary halving of the congestion window on packet loss because of the regular bit errors over the wireless path and hence improves the TCP throughput. When the congestion level of the path is estimated to be high, the threshold for a packet loss to be classified as the congestion loss not to miss any congestion loss is decreased and hence improves the TCP fairness. In ns 2 simulations, the proposed scheme correctly classifies the congestion losses under varying network parameters whereas the previous schemes show some dependency on subsets of parameters.

Comparative performance analysis of versions of TCP in a local network with a lossy link

We use a stochastic model to study the throughput performance of various transport control protocol (TCP) versions (Tahoe (including its older version that we call OldTahoe), Reno, and NewReno) in the presence of random losses on a wireless link in a local network. We model the cyclic evolution of TCP, each cycle starting at the epoch at which recovery starts from the losses in the previous cycle. TCP throughput is computed as the reward rate in a certain Markov renewal-reward process. Our model allows us to study the performance implications of various protocol features, such as fast retransmit and fast recovery. We show the impact of coarse timeouts. In the local network environment the key issue is to avoid a coarse timeout after a loss occurs. We show the effect of reducing the number of duplicate acknowledgements (ACKs) for triggering a fast retransmit. A large coarse timeout granularity seriously affects the performance of TCP, and the various protocol versions differ in their ability to avoid a coarse timeout when random loss occurs; we quantify these differences. We show that, for large packet-loss probabilities, TCP-Reno performs no better, or worse, than TCP-Tahoe. TCP-NewReno is a considerable improvement over TCP-Tahoe, and reducing the fast-retransmit threshold from three to one yields a large gain in throughput; this is similar to one of the modifications in the TCP-Vegas proposal. We explain some of these observations in terms of the variation of fast-recovery probabilities with packet-loss probability. The results of our analysis compare well with a simulation that uses actual TCP code.