Click here for full text:
lmbench3: measuring scalability
Keyword(s): lmbench; benchmarking; distributed systems; parallel systems; multi-processor systems
Abstract: lmbench3 extends the lmbench2 [Staelin98] system to measure a system's performance under scalable load to make it possible to assess parallel and distributed computer performance with the same power and flexibility that lmbench2 brought to uni-processor performance analysis. There is a new timing harness, benchmp, designed to measure performance at specific levels of parallel (simultaneous) load, and most existing benchmarks have been converted to use the new harness. lmbench is a micro-benchmark suite designed to focus attention on the basic building blocks of many common system applications, such as databases, simulations, software development, and networking. It is also designed to make it easy for users to create additional micro-benchmarks that can measure features, algorithms, or subsystems of particular interest to the user.
Back to Index