Efficient Support for Multicomputing on ATM Networks

The emergence of a new generation of networks will dramatically increase the attractiveness of loosely-coupled multicomputers based on workstation clusters. The key to achieving high performance in this environment is efficient network access, because the cost of remote access dictates the granularity of parallelism that can be supported. Thus, in addition to traditional distribution mechanisms such as RPC, workstation clusters should support lightweight communication paradigms for executing parallel applications. This paper describes a simple communication model based on the notion of remote memory access. Applications executing on one host can perform direct memory read or write operations on user-defined remote memory buffers. We have implemented a prototype system based on this model using commercially available workstations and ATM networks. Our prototype uses kernel-based emulation of remote read and write instructions, implemented through unused processor opcodes; thus, applications (or runtime libraries) see direct machine support for remote memory access. We show that this model can be supported safely and efficiently on current systems; for example, a 40-byte remote write operation completes in 30 s on our 140 Mb/s ATM-based network with 19.5 SPECint DECstation hosts. Using example applications, we show how this underlying model can provide excellent performance for both parallel applications and traditional distributed system services.