Six Tips for Selecting HDD and SSD Drives

Gary Watson is Chief Technology Officer, Nexsan, an Imation Company.

g-watson

With today’s wide variety of storage devices comes lots of confusion about what types of drives to use for what data types. Adding to the confusion is Serial ATA (SATA) and SAS, which refer to disk drive interfaces, and Solid State Drive (SSD) which refers to a particular kind of internal technology. Then there are considerations of random access performance, sequential performance, cost, density and reliability.

All these factors make selecting the right drives a challenge. This article offers six tips for navigating through this complexity to help you pick the right solutions for your needs.

1. Don't Confuse Interface Type with Disk Performance or Reliability.

In the past, SAS and SATA were used as convenient shorthand for fast (SAS) or dense (SATA) disk drives. Now, however, we have SSD drives with SATA interfaces as well as inexpensive and dense but relatively low-IOPS 7200 RPM drives with SAS or even Fibre Channel interfaces. Users can no longer make blanket assumptions like “SAS is better for databases.” For example, if we’re comparing a blazing fast SLC SSD with a SATA interface vs. a relatively sluggish 7200 RPM NL-SAS drive, we might be wrong by a factor of 1000x.

Users can’t even use SAS or SATA as shorthand for desired drive reliability. There are several SATA drives that have a claimed 2.0M hour MTBF (mean time between failure), for example a 4TB enterprise hard drive from one of our technology partners. This is in contrast to the typical 1.6M hour MTBF  number for many 3.5-inch 15,000 RPM SAS drives, or the even lower 1.4M hour MTBF number for some 2.5-inch small form factor (SFF) 7200 RPM NL-SAS drives.

Think about that last number for a minute - for a 40TB system, users would need 40 of the 1TB SFF NL-SAS drives, while only needing 10 of the 4TB drives referenced above - one fourth as many. Furthermore, and this is crucial, because the 4TB drive I referenced is so much more reliable, there would be 5 times as many SFF drives failing per year. Additionally, the 4TB drive would only consume 113 Watts, whereas the SFF drives would consume over 200 Watts for the same capacity. When power is a concern, 3.5-inch drive systems often deliver twice the gigabytes per Watt as compared to 2.5-inch drive systems.

2. For Best $/GB, 3.5-inch RPM SATA Is Still King.

Storage vendors have a seemingly endless variety of pricing models, but one constant seems to be that 2.5-inch systems cost twice as much per gigabyte as 3.5-inch systems, assuming both are using “enterprise-grade” drives. But as previously noted, the 3.5-inch solution will be far more reliable.

10K and 15K SAS solutions in either 2.5-inch or 3.5-inch form factor will be approximately 3X to 6X more expensive per gigabyte. SSD solutions can be from 10X to 50X more per gigabyte than comparable SATA drives.

3. HDD Performance is Mostly Dictate by Density and Mechanical Speed.

The random or transactional (IOPS) performance of spinning drives is dominated by the access time, which in turn is determined by rotational latency and seek time. Interface performance has almost no influence on IOPS, except in the negative sense that complex or new interfaces sometimes have bloated or immature driver stacks which can hurt IOPS. Highly random applications which benefit from high IOPS drives include email servers, databases and hypervisor environments.

Sequential performance, which is important for applications like video and D2D backups, are dominated by the RPM of the drive times the bits per cylinder. This number will decrease 50 percent or more as the drive moves from the outermost to the innermost cylinders. Again, as long as the interface is fast enough to keep up (and it is in all modern hard drives), the interface speed (or even the quantity of interface ports) has no measurable effect on sustained performance. The fastest drives today can sustain less than 200 MB/s, which is less than the performance of a single 3 GB SATA port.

4. Consider SSD Instead of 10K OR 15K Drives for Transactional Workloads.

Due to their ever-increasing performance and reliability, 7200 RPM SATA drives are taking on more types of workloads including moderate transactional applications. However, 15,000 RPM drives can deliver roughly two to three times as many small block random transactions as 7200 RPM drives due to their lower rotational latency and much more powerful actuator arm. As a result, they are often used for demanding database or email server workloads.

Recently, SSDs have become mainstream options from most storage vendors. Though not faster at sequential workloads, they are incredibly fast at random small block workloads and may be a superior choice for demanding SQL, Oracle, VMware, Hyper-V and Exchange requirements. Many customers report that they can support more guest virtual machines (VMs) per physical server due to the lower latency of SSD solutions, which may offer tremendous cost savings depending on specifics of licensing and hardware.

SSDs continue to advance at a very fast pace, and are now the leading technology in terms of dollars per IOPS as well as IOPS per watt. Today it is very likely that an all-SSD solution will have lower overall capital and operational cost than one made from 15,000 RPM drives due to the reduction in total slots required to achieve a given transaction performance, and the greatly reduced power footprint as compared to spinning drives for a given number of transactions. Some enterprise SSD’s meet or even exceed the reliability and durability of 15,000 RPM drive systems because far fewer SSD’s are required to achieve any given IOPS level.

5. SSD and 10K/15K Drives Are Not Better for Video.

Somewhat surprisingly, 10,000 RPM and 15,000 RPM drives are not better for video and other media streaming applications, unless there are numerous independent streams being written or read from the same RAID set. In fact, 4TB 7200 RPM drives have higher sequential speed than 3TB drives and often have higher sustained sequential performance than 15,000 RPM drives.

In small RAID sets, the performance limitation might be the drive transfer rate. Therefore, selecting a drive that excels in this area makes sense. In large RAID sets, or with large numbers of drives behind a single controller, the limiting factor is likely to be the RAID engine or the SAN interface technology rather than raw disk speed.  So, drives may be chosen based on other factors such as cost, density or reliability.

Power requirements for 7200 RPM drives are much less than 10,000 RPM or 15,000 RPM drives on a per-GB basis, especially when MAID technology can be used to further reduce power consumption of intermittently or lightly loaded arrays. Video applications often have extended periods of inactivity for some or all of the arrays, so it is an ideal candidate for power savings in this scenario.

6. Sweat the Small Stuff.

Pay close attention to how your vendor treats the subject of hard drives. In the quest for bigger profits, many are moving to a logistics model where the drives are not tested in the storage array until it arrives at the customer site. Others are phasing out the rigorous qualification and ongoing screening process that once was commonplace. Some no longer perform specific qualification checks between drive hardware and firmware revisions, and the hardware and firmware revisions of all the components of the storage array. This is a recipe for disaster, especially with large numbers of drives at a site.

You want to make sure the disk drives you are getting are “enterprise-class,” not consumer grade. Enterprise-class disk drives are those that pass the manufacturer’s highest quality and reliability tests. Frequently new drive technology appears in consumer products before they are released in storage systems designed for data centers, and rightly so. Ask about how the drives have been tested and what quality standards are in place.

Look Past the Hype

There are many considerations when selecting the right drive type for your environment. Don’t let market hype sell you on something just because it’s today’s fad. These six tips can help you sort through some of the confusion. Be clear on your requirements and on the drives, interfaces, price/GB, and finally, reliability available in what you select. Then you are in a better position to make the right choice for your needs.

Industry Perspectives is a content channel at Data Center Knowledge highlighting thought leadership in the data center arena. See our guidelines and submission process for information on participating. View previously published Industry Perspectives in our Knowledge Library.