Comparison and Conclusion
SSD Endurance: Data Integrity is Marathon, Not a SprintThere are multiple measures by which SSDs can be compared. The TBW is one of the newest and best, but there are other factors to consider.
“Do you really want to pick a vendor that may not be around in three years?” asks Seagate’s Teresa Worth. “Especially if you’re a smaller system builder or integrator, do you really want that? Because there’s not enough market to support all the players. With 200 to 300 vendors, there’s going to be consolidation and shakeout. You want a vendor that understands the enterprise and won’t switch products on you frequently, which will force you to redesign, requalify, and increase your costs. You want to make sure they’re going to be around to stand behind their products.”
Another thing to keep in mind is the relationship between duty cycle and warranty. Obviously, it doesn’t make sense to gauge usage by the number of gigabytes written since 200 GB might fill a 200 GB drive while only dent an 800 GB model. Rather, an increasingly common metric is the number of full drive writes conducted in a given time span. Ten drive writes (2 TB of writes on a 200 GB drive, for example) in 24 hours is an increasingly cited figure in enterprise circles. But without a warranty period in which the drive can be expected to reliably repeat this usage, the figure is somewhat meaningless. There is clearly a difference between two drives rated for 10X overwriting per day when one drive has a three-year warranty but the other is covered for five years.
Another thing to watch is how much a given drive uses over-provisioning. Apart from having excess data blocks available to replace failed cells over time, some SSDs will over-provision drives by a given percentage, a practice similar to short stroking a hard drive. This has the effect of increasing both performance and endurance. A 200 GB SSD overprovisioned by 20% will be faster than one without over-provisioning, but the user will only have 160 GB of addressable capacity and the skewed price-per-GB and power efficiency ratios that go with it.
Take a comparison between Intel’s 710 Series SSDs and Seagate’s Pulsar.2. Both are MLC devices targeted at the enterprise. Intel is highly regarded in the SSD world for several good reasons, but Seagate claims more than 3X higher value in terms of performance, endurance, and capacity. This is a bold statement, so let’s break matters down and in so doing learn a bit about how to compare enterprise SSDs.

The above chart shows the impact of over-provisioning, a practice that Intel recommends for its SSDs as a means to increase 710 Series endurance. Now here’s a look at the 710 Series’ performance:
[1] Performance measured using Iometer with 64KB (65,536 bytes) of transfer size with queue depth equal to 32
[1] Performance measured using Iometer with queue depth equal to 32. Measurements performed on full logical block address (LBA) span of drive. Four kilobytes (4KB) equals 4,096 bytes. Eight kilobytes (8KB) equals 8,192 bytes.
[1] Value based on 20 percent over-provisioning.
Let’s compare some numbers. When testing random performance, Iometer shows a 20% over-provisioned 200 GB Intel 710 SSD scoring 38,500 read and 3,300 write IOPS using 4K blocks. (Keep in mind that the user is responsible for over-provisioning Intel’s drive.) Seagate’s Pulsar.2, which is preset with over-provisioning at the factory, scores 48,000 and 15,000 IOPS, respectively—moderately faster on reads but several times faster on random write performance. Keep in mind that Intel is still using the 3 Gb/s SATA interface for its 710 drives while Seagate adopted the 6 Gb/s SATA and SAS interfaces.
With the optional 20% over-provisioning, Intel’s 200 GB 710 drive notes a 1.5 petabyte (PB) total written data endurance. The 200 GB Pulsar.2 more than doubles this, reaching 3.65 PB. An 800 GB Pulsar.2 rates a lifetime endurance of 14.6 PB.
For power efficiency, we need to be careful to reach a common metric across mismatched drive sizes. The way to do this is by examining average operating watts per gigabyte. Intel’s 300 GB 710 SSD yields a power efficiency of 0.0123W/GB. The 800 GB Pulsar.2 yields 0.0074W/GB—up to a 66% W/GB benefit for Seagate and no small concern for data centers deploying hundreds or thousands of drives.
Aside from performance, endurance, power, and warranty (three years for Intel, five for Seagate), Seagate also wins on other enterprise-centric options, such as SAS and SATA Interface support, and TCG-compliant security with 256-bit AES encryption compared to Intel’s 128-bit encryption. (See our prior article on SSD encryption for more about this.)
Both of these enterprise-slanted MLC drives might seem to be similar on the surface, but a dive into their specifics and test results tell a fuller tale. Enterprises need to be deploying the highest value SSDs. That doesn’t mean cheap; it means the drives with the lowest total cost of ownership and highest return on investment. Endurance plays a pivotal role in making that value assessment. Be sure to integrate endurance into your own storage planning and make sure you get a complete picture of the candidates, from independent performance results to JEDEC TBW ratings, before finalizing your strategy.