Problem solve Get help with specific problems with your technologies, process and projects.

Testing RAID for tuning

Testing RAID for tuning

By Rick Cook

It's no secret that the performance of RAID arrays can vary greatly depending on things like block and cache sizes. A number of different factors can affect RAID performance in the real world. This makes it hard to tune a RAID array for best performance.

Lynn Slater, a bay area consultant with experience in large databases, has a presentation on his Web site that discusses characterizing and tuning RAID arrays for best performance. It includes examples as well as a detailed methodology.

Among other things, the presentation discusses the best way to tune a RAID array. Some guidelines: Take one performance area at a time. For example, run a single process and start with tuning buffer and cache tuning on that process first. Then, try different loads on the system. Once single-process performance has been tested, move on to multi-process testing. Then, test with different loads on the storage system.

When testing, it is important to test each area with a range of parameters. To determine the effect of buffer size on performance, test all the buffer sizes with the same-sized file. Then, take one buffer size and vary the file size from very small to maximum. Adjust the cache parameters to see what effect they have on performance.

Before tuning the array, it also helps to understand the surrounding system's average usage requirements. For example, what is the mix of reads and writes on the system? What is the average file size? Once you have the use well-characterized, you can tune the array to maximize performance for that particular mix.

Pay special attention to 'knees', where performance curves flatten off, and to 'hot spots' that generate a disproportionate amount of activity for an increase in load. Generally, there isn't much to be gained by pushing very far beyond the knee of the curve for a particular parameter. A hot spot usually indicates a bottleneck in one or more parameters and you can often get a significant boost in performance by adjusting those parameters at that point.

More resources:

About the author: Rick Cook has been writing about mass storage since the days when the term meant an 80K floppy disk. The computers he learned on used ferrite cores and magnetic drums. For the last twenty years he has been a freelance writer specializing in storage and other computer issues.

Dig Deeper on Primary storage devices

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.