Janardhan R. Iyengar

Concurrent multipath transfer using SCTP multihoming over independent end-to-end paths

Concurrent multipath transfer (CMT) uses the Stream Control Transmission Protocols (SCTP) multihoming feature to distribute data across multiple end-to-end paths in a multihomed SCTP association. We identify three negative side-effects of reordering introduced by CMT that must be managed before efficient parallel transfer can be achieved: (1) unnecessary fast retransmissions by a sender; (2) overly conservative congestion window (cwnd) growth at a sender; and (3) increased ack traffic due to fewer delayed acks by a receiver. We propose three algorithms which augment and/or modify current SCTP to counter these side-effects. Presented with several choices as to where a sender should direct retransmissions of lost data, we propose five retransmission policies for CMT. We demonstrate spurious retransmissions in CMT with all five policies and propose changes to CMT to allow the different policies. CMT is evaluated against AppStripe, which is an idealized application that stripes data over multiple paths using multiple SCTP associations. The different CMT retransmission policies are then evaluated with varied constrained receive buffer sizes. In this foundation work, we operate under the strong assumption that the bottleneck queues on the end-to-end paths used in CMT are independent.

SCTP: a proposed standard for robust Internet data transport

The stream control transmission protocol (SCTP) is an evolving general purpose Internet transport protocol designed to bridge the gap between TCP and UDP. SCTP evolved from a telephony signaling protocol for IP networks and is now a proposed standard with the Internet Engineering Task Force. Like TCP, SCTP provides a reliable, full-duplex connection and mechanisms to control network congestion. However, SCTP expands transport layer possibilities beyond TCP and UDP, offering new delivery options that are particularly desirable for telephony signaling and multimedia applications.

Performance implications of a bounded receive buffer in concurrent multipath transfer

We study the performance of Concurrent Multipath Transfer using SCTP multihoming (CMT) in the presence of a bounded receive buffer (rbuf). We demonstrate using simulation that if two paths are used for CMT, the lower quality (i.e., higher loss rate) path degrades overall throughput of an rbuf-constrained CMT association by blocking the rbuf. We argue that rbuf blocking is not specific to the transport layer, but applies to multipath transfers at other layers as well. We present and discuss CMT performance using several retransmission policies and various constrained rbuf values. We also study the impact of rbuf blocking with different combinations of end-to-end loss rate and delay on the two paths and show that when large differences exist in path delays and loss rates, using only the better path outperforms using two paths concurrently. While rbuf blocking cannot be eliminated, it can be reduced by choice of retransmission policy - a mechanism available to only the transport layer.

Retransmission policies for concurrent multipath transfer using SCTP multihoming

Concurrent multipath transfer (CMT) uses the stream control transmission protocols (SCTPs) multihoming feature to distribute data across multiple end-to-end paths in a multihomed SCTP association. We propose five retransmission policies for CMT. We demonstrate the occurrence of spurious retransmissions in CMT with all of the five policies, and propose an amendment to the timeout retransmission mechanism to avoid spurious retransmissions. We also modify the Cwnd Update for CMT (CUC) algorithm to allow better cwnd growth in CMT with the different retransmission policies. We then evaluate the retransmission policies using ns-2 simulations, and discuss the distributions of traffic that result. We operate under the strong assumptions that the receivers advertised window does not constrain the sender, and that the bottleneck queues on the end-to-end paths used in CMT are independent.

Receive buffer blocking in concurrent multipath transfer

Previously, we studied the performance of concurrent multipath transfer using SCTP multihoming (CMT) under the assumption of an infinite receive buffer (rbuf). Here, we study CMT performance when a sender is constrained by the rbuf. We demonstrate using simulation that if two paths are used for CMT, the lower quality (i.e., higher loss rate) path degrades overall throughput of an rbuf-constrained CMT association by blocking the rbuf. We demonstrate that a wise retransmission policy can alleviate some of the throughput degradation by reducing the rbuf blocking problem. We present and discuss CMT performance using several retransmission policies and constrained rbuf values of 16 KB, 32 KB, 64 KB, 128 KB, and 256 KB. While rbuf blocking cannot be eliminated, it can be reduced by choice of retransmission policy - a facility available to only the transport layer.

SCTP: an innovative transport layer protocol for the web

We propose using the Stream Control Transmission Protocol (SCTP), a recent IETF transport layer protocol, for reliable web transport. Although TCP has traditionally been used, we argue that SCTP better matches the needs of HTTP-based network applications. This position paper discusses SCTP features that address: (i) head-of-line blocking within a single TCP connection, (ii) vulnerability to network failures, and (iii) vulnerability to denial-of-service SYN attacks. We discuss our experience in modifying the Apache server and the Firefox browser to benefit from SCTP, and demonstrate our HTTP over SCTP design via simple experiments. We also discuss the benefits of using SCTP in other web domains through two example scenarios ? multiplexing user requests, and multiplexing resource access. Finally, we highlight several SCTP features that will be valuable to the design and implementation of current HTTP-based client-server applications.

Non-Renegable Selective Acknowledgments (NR-SACKs) for SCTP

In both TCP and SCTP, selectively acked (SACKed) out-of-order data is implicitly renegable; that is, the receiver can later discard SACKed data. The possibility of reneging forces the transport sender to maintain copies of SACKed data in the send buffer until they are cumulatively acked. In this paper, we investigate the situation where all out-of-order data is non-renegable, such as when the data has been delivered to the application, or when the receiver simply never reneges. Using simulations, we show that SACKs result in inevitable send buffer wastage, which increases as frequency of loss events and loss recovery durations increase. We introduce a fundamentally new ack mechanism, Non-Renegable Selective Acknowledgments (NR-SACKs), for SCTP. Using NR-SACKs, an SCTP receiver can explicitly identify some or all out-of-order data as being non-renegable, allowing the sender to free up send buffer sooner than if the data were only SACKed. Simulation comparisons show that NR-SACKs enable efficient utilization of a transport senderpsilas memory. Further investigations show that NR-SACKs also improve throughput in Concurrent Multipath Transfer (CMT) [4].

Retransmission policies with transport layer multihoming

We evaluate several retransmission policies for transport protocols that support multihoming, such as SCTP. We find that schemes that attempt to improve the chance of success by retransmitting to an alternate peer IP address often degrade performance. Our results show that for better performance, new data transmissions and retransmissions should be sent to the same peer IP address. We also find that our multiple fast retransmit algorithm further improves performance by reducing the number of timeouts. Since our results assume reachability of all peer IP addresses, we conclude with suggestions for scenarios where failures are possible. We suggest compromising some of the performance improvements to avoid performance degradation during failures.

Dynamic Window Coupling for multipath congestion control

The traditional problem of end-hosts efficiently and fairly utilizing end-to-end paths becomes significantly harder when the end-hosts are multihomed. Such is the case, for instance, when an end-host has simultaneous connectivity through several service providers, or when a mobile device is simultaneously connected via both a wireless LAN and a cellular network. A multihoming-aware transport protocol, such as MPTCP or SCTP, that sends data over the multiple resulting end-to-end paths must be fair to other flows in the network while being able to maximize its own throughput. In this paper, we present Dynamic Window Coupling (DWC), a multipath congestion control mechanism that seeks to achieve both these goals. DWC uses loss and delay signals to detect shared bottlenecks, explicitly grouping and sharing congestion control across subflows on paths that have a common bottleneck, while separating congestion control for subflows on paths with distinct bottlenecks. DWC detects shifting bottlenecks in the network and responds by dynamically regrouping subflows. Simulations demonstrate that DWC detects shared bottlenecks under most network topologies and conditions that we considered, regroups subflows correctly as bottlenecks shift, aggregates throughput across distinct bottlenecks, and is fair to other TCP flows at all bottlenecks.

Concurrent multipath transfer using SCTP multihoming: introducing the potentially-failed destination state

Previously, we identified the failure-induced receive buffer (rbuf) blocking problem in Concurrent Multipath Transfer using SCTP multihoming (CMT), and proposed CMT with a Potentially-failed destination state (CMTPF) to alleviate rbuf blocking. In this paper, we complete our evaluation of CMT vs. CMT-PF. Using ns-2 simulations we show that CMT-PF performs on par or better than CMT during more aggressive failure detection thresholds than recommended by RFC4960. We also examine whether the modified sender behavior in CMT-PF degrades performance during non-failure scenarios. Our evaluations consider: (i) realistic loss model with symmetric and asymmetric path loss, (ii) varying path RTTs. We find that CMT-PF performs as well as CMT during non-failure scenarios, and interestingly, outperforms CMT when the paths experience asymmetric rbuf blocking conditions. We recommend that CMT be replaced by CMT-PF in future CMT implementations and RFCs.