Here are six tips for selecting HDD and SSD drives that may cause you to think twice.

One of the hottest topics today among IT managers is storage and how to configure it in a virtualized data center that may be complemented by cloud access and requirements for big data applications. Yet is anyone asking the questions—at least the right questions—about what types of hard drives should be used to support these new architectures?

The advent of solid-state drives (SSDs) into the enterprise has introduced a new type of internal technology to storage hardware. The variety of drive types available in the market previously defined only by interface, size, access time, reliability, and, of course, cost is no longer applicable. Is it still appropriate, for example, to say that serial-attached SCSI (SAS) is better for databases when there are very fast SSD drives with serial ATA (SATA) interfaces that can do far better? Sure, we all have heard that SSD drives can be leveraged to serve frequently used data to improve speed. But could the issues be more involved than that? Storage vendor Nexsan has taken on the somewhat perplexing list of questions emerging from today's changing landscape of storage drive hardware in a white paper, "6 Tips for Selecting HDD and SSD Drives."

The company takes into account the different types of environments in which the drives are being asked to perform and makes a number of observations and recommendations useful to those who must decide which drives to deploy in what circumstances. Following are their six tips:

1. Don't confuse interface type with disk performance.

2. For the best cost per gigabyte, 3.5-inch RPM SATA is still king.

3. HDD performance is mostly dictated by density and mechanical speed.

4. Consider SSD instead of 10K or 15K drives for transactional workloads.

5. SSD and 10K/15K drives are not better for video.

6. Sweat the small stuff (don't skip or let your vendor skip testing phases).

Let's take a closer look at each of these and find out exactly what Nexsan has discovered in its in-depth engineering testing.

1. Don't confuse interface type with disk performance. For quite some time, SAS drives were thought of as fast, while SATA drives were generally considered denser. Such shorthand is no longer appropriate, and the two drives are differentiated only by their input/output (I/O) chips. They are physically identical under the covers, according to Nexsan. As such, an "SAS array" filled with 7200 RPM NL-SAS drives will deliver the same performance as ordinary SATA drives of the same family. If someone is looking for a high input/output operations per second (IOPS) configuration for a transactional workload, they may be disappointed by 7200 RPM SAS drives, which deliver the same sequential and random performance as equivalent SATA drives.

Nor should users think of either SAS or SATA drives as being different in terms of reliability, says Nexsan. The typical 3.5-inch 15,000 RPM SAS drive has a typical reliability rating of 1.6M hour mean time between failures (MTBF). The 2.5-inch small form factor (SFF) 7200 RPM NL-SAS drives have typical MTBF ratings of 1.4M hours. (Yes, 3.5-inch drives are generally more reliable than 2.5-inch drives.) Yet, there are several SATA drives with MTBF ratings of 2 million hours, including the 7200 RPM 3TB Hitachi Deskstar 7K3000. So how can you make a blanket statement that SAS drives are more reliable? It's simply not true, says Nexsan.

2.  For the best cost per gigabyte, 3.5-inch 7200 RPM SATA is still king. The tried and true 3.5-inch 7200 RPM SATA drive continues to offer the lowest cost per GB of any common drive. Regardless of the pricing model, users can count on 2.5-inch drive systems costing about twice as much per gigabyte as 3.5-inch systems, assuming both are using enterprise-grade drives. When opting for 10K and 15K SAS solutions in either 2.5-inch or 3.5 inch configurations, expect to pay three to six times more per gigabyte than the slower 7200 RPM drives. SSD solutions can be from 10 to 50 times more expensive per gigabyte than a 3.5-inch 7200 RPM SATA drive.

3. HDD performance is mostly dictated by density and mechanical speed. Rated interface performance almost always has little or no effect on random, or transactional, IOPS, nor on sequential performance. (The exception is complex or new interfaces with immature driver stacks.) The reason is that a single 3GB SATA port is faster than today's fastest drives, which can sustain less than 200BM/s. For random performance, such as against databases, in email servers, or in hypervisor environments, what is important, however, is access time, determined by rotation speed and seek time. Sequential performance, important for video applications and disk-to-disk (D2D) backups, is affected by the RPM of the drive and the bits per cylinder. Speeds will decline dramatically as the drive moves from the outermost to the inner cylinders.

4. Consider SSD instead of 10K or 15K drives for transactional workloads. SSD drives are advancing at a rapid rate, and in terms of cost per IOPS and IOPS per watt, they now are the leader. It's entirely probable that an all-SSD solution will have lower capital and operational costs than one built on 15,000 RPM drives due to the reduction in total slots required to achieve a given transaction performance as well as a reduction in power costs. While concerns about reliability of SSDs have been prevalent, Nexsan says that some enterprise SSDs meet or exceed the reliability of 15,000 RPM mechanical drives. Nexsan prefers single level cell (SLC) over multi level cell (MLC) drives as the company believes they display better random write performance and have the highest write cycle durability of flash-based devices. Other vendors, including IBM, find that MLC SSD drives better meet customer needs than SLC drives. Where high rotation drives might have been considered in the past for demanding database or mail server applications, today SSDs appear to be a better choice.

5.  SSD and 10K/15K drives are not better for video. For video and other streaming media applications, 10,000 and 15,000 RPM drives are not better for video unless there are numerous independent streams being written or read from the same RAID set. A 3TB 7200 RPM drive has a higher sequential speed than does a 2TB drive and often will have a higher sustained sequential performance than the 15,000 RPM drive, according to Nexsan. In small RAID sets, the limiting factor is likely the drive transfer rate. In large RAID sets, or with a single controller supporting a large number of drives, the limiting factor is generally the RAID engine or the SAN interface rather than the disk speed. Therefore, it makes sense to select a drive based on cost or reliability, and you will save on energy with the lower speed drives.

6. Sweat the small stuff. The thrust of this tip is to work with a reliable vendor that cares about testing its drives and doesn't buy seconds or factory rejects. As Nexsan notes, every time a drive is handled or shipped, its reliability goes down. There is a "gray market" in drives, and the products traded by these vendors are generally not coming directly from the factory. A drive considered "enterprise class"— as opposed to "consumer grade"—is a drive that passes the manufacturer's highest quality and reliability tests. Or, as we have heard so often, you get what you pay for.

Knowledge Is Power

It's clear that Nexsan is concerned about storage, but it's also concerned about the customer. Knowing which type of drive to install in which setting can have dramatic implications for system performance, so it's worth learning as much as possible before configuring storage. To download a free copy of the Nexsan white paper from which the above tips were extracted, visit "6 Tips for Selecting HDD and SSD Drives."

Chris Smith was the Senior News Editor at MC Press Online from 2007 to 2012 and was responsible for the news content on the company's Web site. Chris has been writing about the IBM midrange industry since 1992 when he signed on with Duke Communications as West Coast Editor of News 3X/400. With a bachelor's from the University of California at Berkeley, where he majored in English and minored in Journalism, and a master's in Journalism from the University of Colorado, Boulder, Chris later studied computer programming and AS/400 operations at Long Beach City College. An award-winning writer with two Maggie Awards, four business books, and a collection of poetry to his credit, Chris began his newspaper career as a reporter in northern California, later worked as night city editor for the Rocky Mountain News in Denver, and went on to edit a national cable television trade magazine. He was Communications Manager for McDonnell Douglas Corp. in Long Beach, Calif., before it merged with Boeing, and oversaw implementation of the company's first IBM desktop publishing system there. An editor for MC Press Online since 2007, Chris has authored some 300 articles on a broad range of topics surrounding the IBM midrange platform that have appeared in the company's eight industry-leading newsletters. He can be reached at This email address is being protected from spambots. You need JavaScript enabled to view it..