Reiner Ludwig

The Eifel algorithm: making TCP robust against spurious retransmissions

We propose an enhancement to TCPs error recovery scheme, which we call the Eifel algorithm. It eliminates the retransmission ambiguity, thereby solving the problems caused by spurious timeouts and spurious fast retransmits. It can be incrementally deployed as it is backwards compatible and does not change TCPs congestion control semantics. In environments where spurious retransmissions occur frequently, the algorithm can improve the end-to-end throughput by several tens of percent. An exact quantification is, however, highly dependent on the path characteristics over time. The Eifel algorithm finally makes TCP truly wireless-capable without the need for proxies between the end points. Another key novelty is that the Eifel algorithm provides for the implementation of a more optimistic retransmission timer because it reduces the penalty of a spurious timeout to a single (in the common case) spurious retransmission.

Responding to spurious timeouts in TCP

Delays on Internet paths, especially including wireless links, can be highly variable. On the other hand, a current trend for modern TCPs is to deploy a fine-grain retransmission timer with a lower minimum timeout value than 1 s suggested by RFC2988. Spurious TCP timeouts cause unnecessary retransmissions and congestion control back-off. The Eifel algorithm detects spurious TCP timeouts and recovers by restoring the connection state saved before the timeout. This paper presents an enhanced version of the Eifel response to spurious timeouts and illustrates its performance benefits on paths with a high delay-bandwidth product. The refinements concern the following issues (1) an efficient operation in presence of packet losses (2) appropriate restoration of congestion control, and (3) adapting the retransmit timer to avoid further spurious timeouts. In our simulations the Eifel algorithm on paths with a high delay-bandwidth product can increase throughput by up to 250% and at the same decrease the load on the network by 3%. The proposed response also shows adequate performance on heavily congested paths.

Multi-layer tracing of TCP over a reliable wireless link

1. ABSTRACT It is well-known that TCP performance may degrade over paths that include wireless links, where packet losses are often not related to congestion. We examine this problem in the context of the GSM digital cellular network, where the wireless link is protected by a reliable link layer protocol. We propose the use of multi-layer tracing as a powerful methodology to analyze the complex protocol interactions between the layers. Our measurements show that TCP throughput over GSM is mostly ideal and that spurious timeouts are extremely rare. The multi-layer tracing tool we developed allowed us to identify the primary causes of degraded performance: (1) inefficient interactions with TCP/IP header compression, and (2) excessive queuing caused by overbuffered links. We conclude that link layer solutions alone can solve the problem of “TCP over wireless links”. We further argue that it is imperative to deploy active queue management and explicit congestion notification mechanisms in wide-area wireless networks; which we expect will be the bottleneck in a future Internet.

The Eifel retransmission timer

We analyze two alternative retransmission timers for the Transmission Control Protocol (TCP). We first study the retransmission timer of TCP-Lite which is considered to be the current de facto standard for TCP implementations. After revealing four major problems of TCP-Lites retransmission timer, we propose a new timer, named the Eifel retransmission timer, that eliminates these. The strength of our work lies in its hybrid analysis methodology. We develop models of both retransmission timers for the class of network-limited TCP bulk data transfers in steady state. Using those models, we predict the problems of TCP-Lites retransmission timer and develop the Eifel retransmission timer. We then validate our model-based analysis through measurements in a real network that yield the same results.

A future radio-access framework

This paper discusses the requirements on future radio access and, based on the requirements, proposes a framework for such a system. The proposed system based on orthogonal frequency-division multiplexing supports very low latencies and data rates up to 100 Mb/s with wide area coverage and 1 Gb/s with local area coverage. Spectrum flexibility is identified as one main requirement for future systems, and the proposed framework can be deployed in a wide range of spectrum allocations with bandwidths ranging from 2.5 to 100 MHz. Multihop relaying, useful to extend the range for the high data rates, and multiple-antenna configurations are integral parts of the framework. A packet-centric approach is taken for the dataflow processing, implying that the scheduling mechanism and the retransmission protocol operate on complete packets rather than segments thereof, thus allowing for cross-layer optimization. Finally, numerical evaluations are provided, illustrating the feasibility of future very wideband radio access.

Queue management for TCP traffic over 3G links

We propose a novel active queue management scheme tailored to the specific characteristics of third generation (3G) cellular networks. Such links are often the bottleneck for an end-to-end connection and dedicated to one host. Taking advantage of these specific characteristics, we developed a queuing scheme that is simpler than popular random early detection (RED) schemes. Despite its simplicity, our solution yields superior performance for 3G links in terms of high link utilization, low queuing delay, and high end-to-end throughput for TCP bulk data transfers. Our simulation results show a clear win over both conventional drop-tail queuing and also a comparable RED scheme.

An Evolved 3GPP QoS Concept

Simple and cost efficient means for providing and controlling quality of service (QoS) are important for 3GPP operators. This is particularly important for operators planning to provide a wide range of IP-based services across 3GPP broadband shared channels. We point out shortcomings of todays (Rel. 5) 3GPP QoS concept and, based on these, explain why that demand is currently not met. Based on a set of requirements identified for an evolved 3GPP QoS concept, we propose four small additions to the 3GPP specifications. The resulting evolved QoS concept is a realization of DiffServ for 3GPP access networks enhanced with the integration with session admission control. Operator QoS control is exercised from the policy charging rule function (PCRF) and through pre-configuration via the management plane. The evolved QoS concept provides a 3GPP operator with capabilities beyond those found in state-of-the-art fixed broadband access networks

On making SCTP robust to spurious retransmissions

Network anomalies such as packet reordering and delay spikes can result in spurious retransmissions and degrade performance of reliable transport protocols such as TCP and SCTP. Past research showed that in certain networks or paths, such anomalies were fairly common. The result was a series of proposals (Eifel, DSACK-based, F-RTO) targeted towards making a transport robust to such events. Most of the proposals focused on TCP and assumed applicability to SCTP. This paper focuses on SCTP and makes three contributions. First, the effect of spurious retransmissions in SCTP is shown to be more aggravated than in TCP, leading to a problem of congestion window overgrowth. Second, it is shown that most but not all of the solution space is shared between TCP and SCTP. An extension to the Eifel algorithm is proposed. Third, a microscopic as well as macroscopic comparison of the Eifel algorithm and the DSACK-based algorithm is presented. Such a comparison will provide implementers and vendors a clearer view of the applicability and tradeoffs involved in different algorithms.

Lifetime packet discard for efficient real-time transport over cellular links

Mobile cellular users often experience significant delay jitter that undermines quality of real-time applications. Delay jitter can cause unnecessary delivery of stale packets with passed playback deadline and duplicate packets retransmitted by the end host after experiencing a timeout. With Lifetime Packet Discard (LPD) a flow adaptive link can tailor the trade-off between the maximum delay jitter and reliability if quality of service requirements of a flow are known. We propose using an IP option to communicate the flow requirements to the link layer. The packet lifetime is set to the minimum of the data lifetime determined by application and the retransmit timeout value determined by the transport protocol if selective reliability is supported. For congestion-responsive flows, the link transmits only headers of expired packets to prevent unnecessary triggering of end-to-end congestion control. Our simulations show that LPD is efficient in reducing stale data delivery and increases the number of packets delivered in time for real-time flows. For semi-reliable flows throughput and goodput are improved because duplicate packet delivery is prevented.

Link layer analysis of the General Packet Radio Service for GSM

The General Packet Radio Service (GPRS) is one of the services being standardized by ETSI for GSM Phase 2+. This paper briefly presents the status of the GPRS standard giving special attention to the link control/medium access control (RLC/MAC) protocol layer. Subsequently, an analytical evaluation of the GPRS downlink performance is presented. Faced with the increased use of Internet services, the downlink performance is an important metric as users-especially mobile users-will use the Internet mainly for consuming information, i.e. produce downlink traffic load.