Benchmarking Linux With the Phoronix Test Suite
Worth Taking a Look
Knowing how to measure your own computer performance gives you mighty system and network tuning powers. It's also fun to run various benchmarks on commercial products because most of them forbid publishing any kind of benchmark results--but they can't stop you from talking to friends. We're going to take a look at the brand-new Phoronix Test Suite, which is so new the black tape and alligator clips are still visible. The Phoronix Test Suite is for testing hardware performance under Linux. It's still very young and incomplete, but it's worth getting acquainted with--it is based on the the scripts developed by the fine folks (mainly Michael Larabel, it seems) at Phoronix for hardware testing. Phoronix Test Suite is intended to be more than another benchmarking utility; it is an open, extensible platform for creating and customizing all kinds of Linux benchmarking.
Benchmarks are useful , but they're not always precise. They aren't exact matches to real-world use, so it's no good getting all excited over small differences. There is a saying about lies, damned lies, and benchmarks. Presumably the ace Linux administrator is looking for trends, bottlenecks, and what happens when they make changes. What, you say, you mean the purpose isn't to engage in endless arguments over the results, and rig the tests for bigger bragging rights? Well, you can if you want to. I'd rather get useful results.
There are two general categories of benchmarks: synthetic and application. Synthetic benchmarks place an artificial load on a component, like writing and reading a large number of ones and zeros to test CPU and RAM speeds. Application benchmarks measure the performance of real-world applications, such as databases and Web servers. Whatever you use, the more knowledge you have about how the tests operate the more useful your results will be.