Iozone is a highly powerful, open-source filesystem benchmark tool used to analyze a computer’s storage subsystem by testing sequential and random read/write speeds, latency, and throughput. It is widely trusted because it systematically eliminates caching distortions by scaling file sizes past the operating system’s RAM buffers. Core Concepts: The 2-Rule Constraint
Before running Iozone, you must calculate two parameters manually to ensure your benchmarks reflect true disk capabilities instead of fast CPU cache or RAM performance.
File Size (-s) Rule: Set the test file size to at least 2x to 3x the size of your physical RAM. If your system has 16GB of RAM, your file size must be at least 32GB to 48GB. If the file is smaller than your RAM, Linux will cache it, and you will benchmark memory speeds instead of disk speeds.
Record Size (-r) Rule: Set the record block size to match your intended workload. For general testing, 4K or 64K is common. For database servers (like Oracle), match it exactly to the application’s database block size. Key Command Line Options
-a: Full automatic mode. Tests a wide matrix of file and record sizes.
-i #: Runs a specific test type. 0 is write/rewrite, 1 is read/reread, and 2 is random read/write.
-I: Bypasses the OS page cache entirely using Direct I/O (O_DIRECT), forcing physical disk hits.
-t #: Throughput mode. Runs multiple parallel processes or execution threads. -R: Generates standard Excel-compatible output data. Step-by-Step Execution Examples 1. Quick Automatic Standard Benchmark
This option is best if you want Iozone to handle the progression of file sizes up to 512MB. Keep in mind that this is only useful for quick testing, as small file tests will trigger cache hits on modern hardware.
Leave a Reply