The Persistent Relevance of IPC Performance: New Techniques for Reducing the IPC Penalty

Interprocess communication (IPC) is a valuable mechanism for structuring complex systems, as it allows systems to be decomposed into separate protection domains. In addition to being a good structuring mechanism, IPC is a fundamental primitive in distributed and parallel computing; the performance of distributed systems and parallel programs is often determined by the performance of IPC primitives. Recent research has shown how to lower the cost of a round-trip IPC [Bershad et al. 1989, Druschel and Peterson 1992, Johnson and Zwaenepoel 1993, Karger 1989]. For example, a local RPC can be performed in 7.7 μs (254 cycles on a 33 MHz 486) [Yarvin et al. 1993], an RPC across an Ethernet can be performed in 340 μs (8800 cycles on a 25 MHz MIPS) [Thekkath and Levy 1993], and an RPC across an ATM link can be performed in 93 μs (2300 cycles on a 25 MHz MIPS) [Thekkath et al. 1993]. These numbers are remarkable compared to IPC performance in previous systems. Although round-trip IPC time has improved dramatically, the IPC penalty is still significant, particularly in parallel applications and microkernel-based operating systems; we want to reduce the IPC penalty so that it is as close as possible to that of performing a procedure call. In Section 2 we discuss why IPC performance is important. In Section 3 we outline our ideas for reducing the IPC penalty, and in Section 4 we summarize our conclusions.