agsandrew - Fotolia
Spinning disk is dead! Long live spinning disk! It isn't unusual to read both of these headlines in the news nowadays. Of course, as flash prices plummet, it's increasingly common for companies to consider what flash -- either in a hybrid spinning disk-flash combination or in the form of all-flash systems -- might do for workloads, including for hyper-converged storage.
But is 100% hyper-converged flash storage always necessary? Is it possible for companies to stick with the tried-and-true spinning disk in some or all cases?
The state of hyper-converged flash storage
For a couple of years, NAND flash supply was tight, leading to price increases rather than the typical price plummet that accompanies an expanding technology. Since early to mid-2018, though, supplies have loosened, which has led to NAND flash price drops.
Lower prices are relative, however. SSDs on a per-gigabyte basis are still vastly more expensive than their HDD cousins. For example, only six months ago, around the time of the annual Flash Memory Summit, a Wells Fargo analyst noted how enterprise SSDs stood at about a three to four times dollar-per-gigabyte premium compared to mission-critical HDDs, while SSDs were at about a 15 to 17 times dollar-per-gigabyte premium compared to nearline, high-capacity HDDs for the enterprise.
That's still a hefty delta, and even though workloads might like it, not all workloads demand all-flash. So, the question is: If you're considering hyper-converged infrastructure (HCI), why might you consider an all-disk configuration and not hyper-converged flash storage only?
Capacity vs. performance
The answer is easy: Consider all-disk over all-flash or hybrid flash for hyper-converged storage and HCI platforms when you value capacity over performance to a sufficient extent. That line is incredibly subjective, of course, and will vary based on a number of factors. These include an organization's tolerance for additional costs to ensure application performance and the application mix that is being supported by the hyper-converged storage platform.
If you're operating an application that is especially write-heavy, you may need to consider hyper-converged flash storage, at least in a hybrid way. But, again, it depends on the sustained write needs of the application. Analyze each of the applications you intend to run in the environment to determine what's needed.
Dig Deeper on Hyper-Converged Infrastructure Systems
Related Q&A from Scott D. Lowe
Storage for IoT devices and applications requires new strategies and approaches. Find out how three approaches are being used to facilitate storage ... Continue Reading
Cloud database services, object storage services and application memory cache are among the options available for storage in serverless environments. Continue Reading
AI-based analytics capabilities are becoming pervasive in storage arrays. Find out how they're helping get the unstructured data deluge under control. Continue Reading