The ERICA switch algorithm for ABR traffic management in ATM networks

Abstract

We propose an explicit rate indication scheme for congestion avoidance in ATM networks. In this scheme, the network switches monitor their load on each link, determining a load factor, the available capacity, and the number of currently active virtual channels. This information is used to advise the sources about the rates at which they should transmit. The algorithm is designed to achieve e ciency, fairness, controlled queueing delays, and fast transient response. The algorithm is also robust to measurement errors caused due to variation in ABR demand and capacity. We present performance analysis of the scheme using both analytical arguments and simulation results. The scheme is being implemented by several ATM switch manufacturers. 1We begin by brie y examining the ABR service. In section 3, we describe basic concepts such as the switch model and design goals. Section 4 describes the algorithm. Section 5 presents representative simulations to show that the scheme works under stressful conditions; we also present analytical arguments in appendix A. Finally, our conclusions are presented in section 6.2 The ABR Control Mechanism ATM networks o er ve classes of service: constant bit rate CBR, real-time variable bit rate rt-VBR, non-real time variable bit rate nrt-VBR, available bit rate ABR, and unspeci ed bit rate UBR. Of these, ABR and UBR are designed for data tra c, which has a bursty unpredictable behavior.The UBR service is simple in the sense that users negotiate only their peak cell rates PCR when setting up the connection. If many sources send tra c at the same time, the total tra c at a switch may exceed the output capacity causing delays, bu er over ows, and loss. The network tries to minimize the delay and loss using intelligent bu er allocation 15 , cell drop 16 and scheduling, but makes no guarantees to the application.The ABR service provides better service for data tra c by periodically advising sources about the rate at which they should be transmitting. The switches monitor their load, compute the available bandwidth and divide it fairly among active o ws. This allows competing sources to get a fair share of the bandwidth and not be starved due to a small set of rogue sources. The feedback from the switches to the sources is sent in Resource Management RM cells which are sent periodically by the sources and turned around by the destinations see gure 1.The RM cells contain the source current cell rate CCR, and several other elds that can be used by the switches to provide feedback to the source. These elds are: Explicit Rate ER, Congestion Indication CI Flag, and No Increase NI Flag. The ER eld indicates the rate that the network can support at the particular instant in time. When starting at the source, the ER eld is usually set to the PCR, and the CI and NI ags are clear. On the path, each switch reduces the ER eld to the maximum rate it can support and sets CI or NI if necessary 12 .The RM cells owing from the source to the destination are called forward RM cells FRMs while those

Architecture for dynamic and fair distribution of bandwidth

“…Unlike the ECN extension, the network feedback in the X-BDS model not only notifies the edge routers about congestion but also carries additional information such as the arrival rate and aggregate RBR on the congested link. A similar idea is used in ATM networks for Available Bit Rate (ABR) congestion control [24] where the feedback carried by the resource management cells also includes rate information. However, the ABR congestion control relies on per-flow information stored in the network core and tries to achieve utilization goals first and only then seeks fairness.…”

Optimal structured feedback policies for ABR flow control using two-timescale SPSA

“…In what follows, we refer to our main policy simply as Policy (13) for convenience. Next, let us briefly explain the ERICA algorithm [16]. ERICA computes the available bandwidth ABR according to ABR Target utilization (Link Capacity (VBR CBR) ).…”

“…These schemes were found to substantially improve system performance and offer more flexibility to switch designers. Several explicit-rate feedback schemes have been proposed in recent years; amongst these are Enhanced PRCA (EPRCA) [22], the Dynamic Max Rate Control Algorithm (DMRCA) [10], and the well-known Explicit Rate Indication for Congestion Avoidance (ERICA) [16]. We shall consider the last algorithm above in detail for comparison purposes with our algorithm.…”

Fair flow control for ATM-ABR multipoint connections

“…It is not essential if its implementation complexity is high. In [15], we give a proof that this algorithm converges to the max-min fair rates for a single bottleneck case.…”

“…A dynamic queue control function is used to scale the available capacity and achieve a constant queuing delay in steady state [15]. The "target delay" parameter specifies the desired queuing delay.…”