RAID arrays with dozens of hard drives are not uncommon for reaching certain performance levels. We demonstrate how beautifully SSD RAID arrays can scale. There may come a time when a few flash-based drives will replace entire farms of hard disks.
This article originally appeared here on Tom’s Hardware. It has been edited for the busy IT pro. Please see the original for details excluded here and for information on how to buy products featured herein.
Barely a week goes by without a new product being introduced to the growing SSD market. Meanwhile, the storage landscape is already packed with MLC and SLC NAND-based solid state drives claiming superlative data throughput rates of more than 250 MB/s (on SATA 3 Gb/s ports) and I/O rates in the five-figure range. In contrast, the veteran hard drives seem like relics from a seemingly bygone era: cheap, much slower, and eventually doomed.
It is not really quite that simple, of course, because when the underlying technology does not fit, the SSD performance data that looks so impressive on paper can quickly go up in smoke, and even fall behind that of notebook hard drives. A flash drive can only reach its full potential with the right combination of hardware resources, controller, cache, and software features.
But that is only a basic requirement and you have to consider other factors, including the latest Serial ATA drivers and SSD firmware, AHCI support through BIOS, as well as the TRIM feature offered in Windows 7, Windows Server 2008 R2, and Linux distributions with kernel version 2.6.18 or higher. This keeps the SSD informed of deleted blocks so that the available storage space is managed better, thereby preventing performance degradation.
Flash Drives For Corporate Use
Because of their technology, SSDs are not only interesting for PC enthusiasts and performance aficionados, but for corporate use as well. Regardless of the usage scenario, there are a lot of technical advantages that favor SSDs: while only a minority of enterprise sector users will benefit a lot from the great throughput rates, the lack of moving parts means superior access times, as well as lower operating temperatures. And most important, especially in servers dealing with huge numbers of individual read and write operations, the I/O performance is miles away from traditional hard drives.
The few drawbacks of SSDs are easy to list: the price per gigabyte is still much higher than for traditional hard drives. Also, the lifetime of flash memory is technically limited to a certain number of write cycles. This is not usually a significant disadvantage, especially since this problem is shared with traditional hard drives due to mechanical wear and tear, and the latest high-end flash products have a life expectancy purported to be on par with enterprise hard drives.
How Do Enterprise SSDs Scale In RAID?
Under what conditions is the use of SSDs worth the investment for a company? We address that question in this article, and answer it from two angles. First, we will investigate the scenarios where the use of SSDs is worth the investment over traditional enterprise hard drives. And because the RAID topic inevitably surfaces in this context, we also take a look at SSD RAID scalability.
The fact that an SSD RAID array almost always dominates a comparable hard drive RAID array in terms of performance has been extensively studied by us and others, and is therefore not the main focus of this article. Instead, what is important here is whether the 'Online Capacity Expansion' feature of RAID arrays now also resembles 'Online I/O Capability Expansion,' because the I/O performance increases significantly with every drive added, and it is higher than with traditional hard disks.
Achim Roos is Test Editor at Tom's Hardware Germany. He is a certified engineer who develops benchmarks and performs testing on CPUs, all kinds of storage, and platforms