How to Test RAID Speed

You could test your RAID's speed by timing how long it takes to copy a particular file to the array. Most operating systems display a progress bar while large file operations are taking place, allowing you to see roughly how long the process will take. This is useful if you just want to estimate your RAID's performance, as you do not have to install any extra software. However, this method is not generally as accurate as a benchmarking test.

Using a benchmarking tool gives you greater control over your speed tests, enabling you to vary parameters, such as the block size used by the array and how long the test runs. Using a small block size lets you test the RAID under stressful conditions, forcing the array to undertake more input/output operations than normal to write the data as a series of small data blocks rather than fewer larger blocks. Running a long or continuous test helps you to see how the array would cope with periods of extended use.

Benchmarking applications vary in their scope and intended purpose. Some, such as Atto's Disk Benchmark and Intel's IOmeter, are specifically intended to test RAID applications. Disk Benchmark can test the performance of individual drives within a RAID array as well as the controller card to which the array is connected. However, hard disk testing tools such as HDTune also work with RAID volumes, testing each volume as if it were one big hard drive. Such applications usually provide a simpler interface at the expense of reduced functionality.

One of the key factors affecting RAID speed is the RAID mode used. Different RAID modes combine drives together in different ways, with some geared toward performance and some toward reliability. RAID controller card speed is also important, with faster cards able to transfer more data to the RAID in a given space of time. In addition, you should consider the spindle speed of the drives used in your array. Enterprise-grade drives usually spin at 7200 RPM, producing significantly faster results than 5400 RPM consumer drives.