Adaptive Routing on the New Switch Chip for IBM SP Systems

Abstract

The IBM RSÂ6000 SP is one of the most successful commercially available multicomputers. SP owes its success partially to the scalable, high bandwidth, low latency network. This paper describes the architecture of Switch2 switch chip, the recently developed third generation switching element which future IBM RSÂ6000 SP systems may be based on. Switch2 offers significant enhancements over the existing SP switch chips by incorporating advances in both VLSI technology and interconnection network research. One of the major new features of Switch2 is the incorporation of adaptive routing support into it. We describe the adaptive source routing architecture of the Switch2 chip which is a unique feature of this chip. The performance of the adaptive source routing and oblivious routing for a wide range of system characteristics and traffic patterns is evaluated. It is shown that adaptive source routing outperforms or performs comparably with oblivious routing. We propose two novel algorithms for generating adaptive routes specifications required for enabling the usage of adaptive source routing. A comparison between the cost of these two algorithms and the performance improvement obtained from using these algorithms are discussed. We also propose different output selection functions to be used in switching elements for implementing the adaptive routing. We evaluate and compare the performance of these selection functions and discover that the best selection functions for BMINs are not dependent on the traffic pattern, message size, or system size. Academic Press

Deterministic versus Adaptive Routing in Fat-Trees

“…n the number of nodes in the network 1 . This interval is valid for all the ascending links of the switch and can result in a cyclic interval.…”

“…Remember that FIR implements the routing function. In [1], the authors compare several selection functions for adaptive routing, but they only consider the peak throughput in the evaluation, so they conclude that selection function is not critical. However, in [6], several selection functions for fat-trees were proposed and the authors notice that selection function has a high impact over average latency.…”

“…The authors in [1] compare an oblivious routing algorithm with an adaptive routing algorithm for multistage networks, and conclude that an adaptive algorithm achieves a higher performance. The readers should take into account that oblivious routing is not the same than deterministic routing [4,3], since oblivious routing can provide several paths for a source-destination pair and the routing decision is taken without considering (oblivious to) network status.…”

Adaptive Routing in InfiniBand Hardware

“…A large body of work has shown that AR will offer superior performance compared to deterministic and oblivious routing for many non-uniform traffic scenarios. The evaluation work herein has mainly been conducted as simulation studies [1], [9], [14], [17], [18], [24], [25], [28]. On the contrary, there are also contributions revealing that deterministic routing (DR) for some scenarios and topologies, such as Fat trees, achieve a similar, or even a higher throughput than AR.…”

Low-Memory Techniques for Routing and Fault-Tolerance on the Fat-Tree Topology

“…In source routing, the source node computes the path prior to packet injection and stores it in the packet header, which may noticeably increase the packet header length. Source routing has been used in some networks because routers are very simple [8,15]. In this routing scheme, the switches of the network just read the stored route in the packet header and forward the packet through the indicated output port, which makes routing very quickly and switches very simple.…”