With almost every technology come myths, stories told to scare off potential new customers or, sometimes, to help them understand the challenges that they may face when jumping into the water. In this way, flash storage is no different than any other technology out there. But, as is always the case, myths, while once rooted in fact, may no longer apply or may be exaggerated for effect. In this post, I’ll debunk five common myths that surround flash storage technology, or at least explain why the myth doesn’t quite match the reality.
Myth #1: Solid state storage isn’t reliable enough for enterprise use
It’s true that the individual cells that comprise solid state storage devices have a finite number of read/write cycles before they become unusable. It’s also true that applications that continually thrash the same cells over and over will accelerate the demise of these cells. However, the problem is not as intractable as it would first appear.
The solid state storage devices on the market today implement various wear leveling techniques designed to ensure that host systems don’t simply batter the same cells over and over again, but instead spread the workload across all of the cells available on the system. This technique significantly improves the overall life of the disk to a point at which it will likely outlast any service life for which it is recruited.
It’s important to note that different kinds of flash chips that comprise solid state disks carry different levels of reliability:
- Single Level Cell (SLC): The most expensive kind of flash storage is also the most reliable and fastest, with cells that can with stand 100,000 program/erase cycles per cell.
- Multi-Level Cell (MLC): A common flash storage type used in business storage, MLC can generally withstand up to 10,000 program/erase cycles per cell
- Enterprise MLC (eMLC): Also commonly used in enterprise storage, eMLC is more robust than MLC and often carries the ability to withstand up to 30,000 program/erase cycles per cell.
- Triple Level Cell (TLC): TLC is an emerging, very low cost way to get solid state storage, but it comes at the cost of performance and reliability, with TLC only able to withstand 5,000 program/erase cycles per cell.
While 100,000 program/erase cycles may not sound like a lot, bear in mind that wear leveling spreads this load, so that 100,000 cycles can go a very long way.
Myth #2: Solid state storage is too expensive when compared to hard drive storage
Yes. In terms of dollars per GB, solid state storage is more expensive than hard disk-based storage. flash storage pricing continues to drop, when it’s possible to pick up a 4 TB hard drive at New Egg for less than $200, flash storage will not be winning the capacity-based price war anytime in the near future.
However, for organizations suffering from major storage-induced performance issues, cost based on performance as opposed to capacity may point to flash storage being on par with or even less expensive than rotational disks. Whereas just about any traditional hard drive tops out at around 200 IOPS, a single eMLC-based solid state disk may support 10,000 or more IOPS.
For this myth, if the primary metric is direct capacity on a disk-by-disk basis, then the SSD would, in fact, be much more expensive, but when it comes to performance, SSDs can often deliver speeds that spinning disk cannot come close to matching, which can save money by increasing efficiency (reducing latency) and in some cases, provide an important competitive advantage.
Myth #3: Solid state storage IOPS is the only metric of importance
To prove that solid state disks are really fast – as I did in the previous myth – the measure of IOPS is often used. However, regardless of disk type, IOPS is only one part of the larger story. In fact, one metric is generally considered much more important than IOPS –latency. This is the amount of time that it takes for a storage command from a host to be satisfied by the underlying storage. The higher the latency value, the longer the host sits awaiting a response, and the longer the user waits to get work done.
Besides, IOPS numbers can be pretty easily manipulated to look really good in marketing materials. This is true due to the number of variables present in an IOPS calculation. With latency, the result is based on some simple math, so it’s easier to do comparisons.
Myth #4: Solid state storage isn’t a fit for the SMB market
I’ve heard people say that solid state is an upmarket play. Unfortunately, that’s not reality, either. Solid state storage is just as applicable to the SMB as it is to the enterprise; it’s all about the applications that will be run in the environment. An SMB trying to do VDI with even 100 machines may find better success using solid state storage than rotational. An SMB working with big data sets may find better success using SSDs than rotating disks. When it comes down to the market, solid state is targeted more at applications than it is at specific market segments.
Myth #5: Solid state storage arrays don’t have enough capacity for most customers
It’s true that solid state disks don’t come in jumbo size, like their rotational cousins. However, it is possible to get 1 TB SSDs, and many storage vendors are finding ways to make SSDs more attractive from a capacity standpoint. For example, companies such as Tegile have baked in powerful deduplication and thin provisioning capabilities into their arrays which, when combined, the company claims provides a significant capacity improvement over the raw size of the array.
In addition, for those companies that need a lot of capacity but that need more performance than can be had with rotating disks alone, there is a burgeoning hybrid market with such players as Tegile. In these systems, solid state storage is leveraged as a mega-cache for large capacity rotating disks. As such, these arrays are capable of delivering tens or hundreds of thousands of IOPS while also providing up to 500 terabytes of capacity, when combined with data reduction technologies.
Action Item: Don’t believe everything you read! Before falling for a myth, check it out and make sure that the conditions that originally spawned the myth haven’t changed and that you’re looking at the market with the right perspective.