lmbench

lmbench is a portable suite of microbenchmarks for Unix-like operating systems, designed to measure fundamental system performance. It focuses on low-level latency and bandwidth characteristics of operating-system primitives and hardware, such as process creation, context switches, interprocess communication, memory access patterns, file I/O, and network communication. By providing small, repeatable measurements, lmbench helps researchers and performance engineers compare systems and kernel configurations under controlled conditions.

Tests cover areas including fork and exec latency, context-switch time between processes, IPC latency via pipes, Unix domain sockets and shared memory, system-call overhead, memory latency across cache levels, memory bandwidth, page fault timing, and file system latency. Network tests measure round-trip latency and throughput using TCP between machines. The suite emphasizes reproducibility and simple interpretation of results rather than simulating full applications.

lmbench is designed to be portable across Unix-like systems and typically runs on Linux, BSD variants, Solaris, and macOS. It outputs results in plain text, making it easy to parse and graph. The tests use small data sets and repeated runs to obtain stable estimates in microseconds and megabytes per second.

Originating in the 1990s, lmbench has become a widely cited tool for low-level system performance evaluation and kernel research. It is distributed as open-source software and remains in use in academic and industry settings to compare hardware platforms, kernel revisions, and configuration options. As with any microbenchmark, results should be interpreted with care, as real-world workloads may behave differently from synthetic tests.