RAID Scaling Charts, Part 1

RAID Testing With 2-8 Drives

Most enthusiast and mainstream users would consider setting up a RAID array mainly for performance reasons - few really care about data safety. For this reason, the majority of arrays installed consists of only two drives, which run a simple RAID 0 stripe set. Haven’t you ever asked yourself how these RAID arrays scale as you increase the number of hard drives? Is it worth it? When does the array hit bottlenecks? Administrators and über users be ready: here come the ultimate Tom’s Hardware RAID Scaling Charts.

This is the first of two articles. We will discuss RAID 0, RAID 1 and RAID 0+1 here, while part 2 will deal with RAID 5 and RAID 6.

Why RAID Charts?

There are many reasons why we decided to carry out this project. On the one hand, there are numerous applications that require high storage performance: think of audio or video applications that rely on a constant data stream, such as high resolution HD video recording or audio recording. As soon as you reduce or drop compression your data rates will skyrocket; they also will with multiple recording channels. File servers in large networks also need to be able to provide software updates in a fast manner to multiple client PCs at the same time. Or think of data collection servers that store dozens of surveillance camera photos per second, communication logs, or data coming from sensors within a particle accelerator or other scientific application.

Even so, high throughput isn’t everything: some applications require hundreds of I/O operations per second. Large databases that cannot be cached into system memory rely on high storage I/O performance as well. Here, you might think of wire transfers, credit card transactions, orders that affect inventory control systems, and so forth.

Our RAID charts include all the results that are important to find a good match between your application requirements and the number of hard drives you’ll need. We’ve included access time measurements, interface bandwidth performance, maximum, and average and minimum read and write transfer rates. We also include I/O performance benchmarks for typical applications such as file servers, web servers, database servers or workstations.

Join our discussion on this topic

Patrick Schmid
Editor-in-Chief (2005-2006)

Patrick Schmid was the editor-in-chief for Tom's Hardware from 2005 to 2006. He wrote numerous articles on a wide range of hardware topics, including storage, CPUs, and system builds.