Using multiple per egress burstifiers for enhanced TCP performance in OBS networks

Burst assembly mechanism is one of the fundamental factors that determine the performance of an optical burst switching (OBS) network. In this paper, we investigate the influence of the number of burstifiers on TCP performance for an OBS network. The goodput of TCP flows between an ingress node and an egress node traveling through an optical network is studied as the number of assembly buffers per destination varies. First, the burst-length independent losses resulting from the contention in the core OBS network using a non-void-filling burst scheduling algorithm, e.g., Horizon, are studied. Then, burst-length dependent losses arising as a result of void-filling scheduling algorithms, e.g., LAUC-VF, are studied for two different TCP flow models: FTP-type long-lived flows and variable size short-lived flows. Simulation results show that for both types of scheduling algorithms, both types of TCP flow models, and different TCP versions (Reno, Newreno and Sack), TCP goodput increases as the number of burst assemblers per egress node is increased for an OBS network employing timer-based assembly algorithm. The improvement from one burstifier to moderate number of burst assemblers is significant (15–50% depending on the burst loss probability, per-hop processing delay, and the TCP version), but the goodput difference between moderate number of buffers and per-flow aggregation is relatively small, implying that an OBS edge switch should use moderate number of assembly buffers per destination for enhanced TCP performance without substantially increasing the hardware complexity.

TCP Window-Based Flow-Oriented Dynamic Assembly Algorithm for OBS Networks

In transport control protocol (TCP) over optical burst switching (OBS) networks, TCP window size and OBS parameters, including assembly period and burst dropping probability, will impact the network performance. In this paper, a parameter window data dropping probability(WDDP), is defined to analyze the impact of the assembly and the burst loss on the network performance in terms of the round trip time and the throughput. To reduce the WDDP without introducing the extra assembly delay penalty, we propose a novel TCP window based flow-oriented assembly algorithm dynamic assembly period (DAP). In the traditional OBS assembly algorithms, the packets with the same destination and class of service (CoS) are assembled into the same burst, i.e., the packets from different sources will be assembled into one burst. In that case, one burst loss will influence multiple TCP sources. In DAP, the packets from one TCP connection are assembled into bursts, which can avoid the above situation. Through comparing the two consecutive burst lengths, DAP can track the variation of TCP window dynamically and update the assembly period for the next assembly. In addition, the ingress node architecture for the flow-oriented assembly is designed. The performance of DAP is evaluated and compared with that of fixed assembly period (FAP) over a single TCP connection and multiple TCP connections. The results show that DAP performs better than FAP at almost the whole range of burst dropping probability.     

Reducing the impact of false time out on TCP performance in TCP over OBS networks

Random burst contention losses plague the performance of Optical Burst Switched networks. Such random losses occur even in low load network condition due to the analogous behavior of wavelength and routing algorithms. Since a burst may carry many packets from many TCP sources, its loss can trick the TCP sources to conclude/infer that the underlying (optical) network is congested. Accordingly, TCP reduces sending rate and switches over to either fast retransmission or slow start state. This reaction by TCP is uncalled-for in TCP over OBS networks as the optical network may not be congested during such random burst contention losses. Hence, these losses are to be addressed in order to improve the performance of TCP over OBS networks. Existing work in the literature achieves the above laid objective at the cost of violating the semantics of OBS and/or TCP. Several other works make delay inducing assumptions. In our work, we introduce a new layer, called Adaptation Layer, in between TCP and OBS layers. This layer uses burst retransmission to mitigate the effect of burst loss due to contention on TCP by leveraging the difference between round trip times of TCP and OBS. We achieve our objective with the added advantage of maintaining the semantics of the layers intact.     

TCP over optical burst-switched networks with controlled burst retransmission

For optical burst-switched (OBS) networks in which TCP is implemented at a higher layer, the loss of bursts can lead to serious degradation of TCP performance. Due to the bufferless nature of OBS, random burst losses may occur, even at low traffic loads. Consequently, these random burst losses may be mistakenly interpreted by the TCP layer as congestion in the network. The TCP sender will then trigger congestion control mechanisms, thereby reducing TCP throughput unnecessarily. In this paper, we introduce a controlled retransmission scheme in which the bursts lost due to contention in the OBS network are retransmitted at the OBS layer. The OBS retransmission scheme can reduce the burst loss probability in the OBS core network. Also, the OBS retransmission scheme can reduce the probability that the TCP layer falsely detects congestion, thereby improving the TCP throughput. We develop an analytical model for evaluating the burst loss probability in an OBS network that uses a retransmission scheme, and we also analyze TCP throughput when the OBS layer implements burst retransmission. We develop a simulation model to validate the analytical results. Simulation and analytical results show that an OBS layer with controlled burst retransmission provides up to two to three orders of magnitude improvement in TCP throughput over an OBS layer without burst retransmission. This significant improvement is primarily because the TCP layer triggers fewer time-outs when the OBS retransmission scheme is used.     

A stochastic model for the behavior of multiple TCP NewReno sources over optical burst switching network

This work proposes a stochastic model to characterize the transmission control protocol (TCP) over optical burst switching (OBS) networks which helps to understand the interaction between the congestion control mechanism of TCP and the characteristic bursty losses in the OBS network. We derive the steady-state throughput of a TCP NewReno source by modeling it as a Markov chain and the OBS network as an open queueing network with rejection blocking. We model all the phases in the evolution of TCP congestion window and evaluate the number of packets sent and time spent in different states of TCP. We model the mixed assembly process, burst assembler and disassembler modules, and the core network using queueing theory and compute the burst loss probability and end-to-end delay in the network. We derive expression for the throughput of a TCP source by solving the models developed for the source and the network with a set of fixed-point equations. To evaluate the impact of a burst loss on each TCP flow accurately, we define the burst as a composition of per-flow-bursts (which is a burst of packets from a single source). Analytical and simulation results validate the model and highlight the importance of accounting for individual phases in the evolution of TCP congestion window.     

Performance evaluation of high speed TCP over optical burst switching networks

Study of TCP performance over OBS networks has been an important problem of research lately and it was found that due to the congestion control mechanism of TCP and the inherent bursty losses in the Optical Burst Switching (OBS) network, the throughput of TCP connections degrade. On the other hand, High Speed TCP (HSTCP) was proposed as an alternative to the use of TCP in high bandwidth-delay product networks. HSTCP aggressively increases the congestion window used in TCP, when the available bandwidth is high and decreases the window cautiously in response to a congestion event. In this work, we make a thorough simulation study of HSTCP over OBS networks. While the earlier works in the literature used a linear chain of nodes as the network topology for the simulation, we use the popular 14-node NSFNET topology that represents an arbitrary mesh network in our study. We also study the performance of HSTCP over OBS for different bandwidths of access networks. We use two different cases for simulations where in the first HSTCP connections are routed on disjoint paths while in the second they contend for resources in the network links. These cases of simulations along with the mesh topology help us clearly distinguish between the congestion and contention losses in the OBS network and their effect on HSTCP throughput. For completeness of study, we also simulate TCP traffic over OBS networks in all these cases and compare its throughput with that of HSTCP. We observe that irrespective of the access network bandwidth and the burst loss rate in the network, HSTCP outperforms TCP in terms of the throughput and robustness against multiple burst losses up to the expected theoretical burst loss probability of 10⁻³.     

Traffic statistics and performance evaluation in optical burst switched networks

Optical burst switching (OBS) is a promising switching technology to exploit the potential benefits of optical communication and, at the same time, support statistical multiplexing of data traffic at a fine granularity. To quantify its benefits, the paper describes several typical burst assembly algorithms and studies their impact on the assembled burst traffic characteristics as well as the performance of TCP traffic. Also described is a proactive burst scheduling algorithm, called burst overlap reduction algorithm (BORA), which schedules locally assembled bursts in such a way as to reduce burst contention at downstream nodes in OBS networks. Furthermore, to provide analytical insights into performance evaluation of OBS networks, a burst loss model at an OBS node and its extension to different reservation protocols are presented.     

Improving Reno and New-Reno's performances over OBS networks without SACK

In TCP over optical burst switching (OBS) networks, consecutive multiple packet losses are common since an optical burst usually contains a number of consecutive packets from the same TCP sender. It has been proved that over OBS networks Reno and New-Reno achieve lower throughput performances than that of SACK, which can address the inefficiency of Reno and New-Reno in dealing with consecutive multiple packet losses. However, SACK adopts complex mechanisms not only at the sender's but also at the receiver's protocol stack, and thus has a higher difficulty in deployment.In this paper we propose B-Reno, a new TCP implementation designed for TCP over OBS networks. Using some simple modifications to New-Reno only at the sender's protocol stack, B-Reno can overcome the inefficiencies of Reno and New-Reno in dealing with consecutive multiple packet losses and thus improve their throughputs over OBS networks. Moreover, B-Reno can also achieve performance similar with that of SACK over OBS networks while avoiding SACK's difficulty in deployment due to complex mechanisms at both the sender's and the receiver's protocol stack.     

Congestion window-based adaptive burst assembly for TCP traffic in OBS networks

Burst assembly is one of the key factors affecting the TCP performance in optical burst switching (OBS) networks. When the TCP congestion window is small, the fixed-delay burst assembler waits unnecessarily long, which increases the end-to-end delay and thus decreases the TCP goodput. On the other hand, when the TCP congestion window becomes larger, the fixed-delay burst assembler may unnecessarily generate a large number of small-sized bursts, which increases the overhead and decreases the correlation gain, resulting in a reduction in the TCP goodput. In this paper, we propose adaptive burst assembly algorithms that use the congestion window sizes of TCP flows. Using simulations, we show that the usage of the congestion window size in the burst assembly algorithm significantly improves the TCP goodput (by up to 38.4% on the average and by up to 173.89% for individual flows) compared with the timer-based assembly, even when the timer-based assembler uses the optimum assembly period. It is shown through simulations that even when estimated values of the congestion window size, that are obtained via passive measurements, are used, TCP goodput improvements are still close to the results obtained by using exact values of the congestion window.