Analyzing communication latency using the Nectar communication processor
Peter Steenkiste · 1992
Abstract Reducing latency has traditionally been the biggest For multicomputer applications, the most important challenge. Dedicated multicomputers with special-purpose performance parameters of a network is the latency for short messages. In this paper we present an analysis of communication latency using measurement of the Nectar system. Nectar is a high-performance interconnects such as the Intel Touchstone system have latencies below 100 microseconds [4], while latencies between Unix workstations communicating over general multicomputer built around a high-bandwidth networks are typically one order of magnitude higher. The crosspoint network. Nodes are connected to the Nectar latency between two Sun4/330 running Sun OS 4.1 is for network using network coprocessors that are primarily responsible the protocol processing, but that can also execute application code. This architecture allows us to analyze message latency both between workstations example about 800 microseconds. One reason for the higher latency is the difference in communication medium. The interconnection networks used by dedicated with an outboard protocol engine and between multicomputers only have to cover a few meters and lightweight nodes with a minimal runtime system and a guarantee data integrity in hardware, while general fast, simple network interface (the coprocessors). We study how much context switching, buffer management and protocol processing contribute to the communication latency and we discuss how the latency is influenced by the protocol implementation. networks have to cover much larger distances and introduce errors in the data stream with non-zero probability. The communication protocols that recover from these errors introduce overhead, thus adding to the latency. This overhead however is, or should be, of the 1.