Nexsan recently announced the release of its new E-Series 18F (E18F) storage platform. I had the chance to chat with Surya Varanasi, CTO of StorCentric, about the announcement and thought I’d share some thoughts here.
Less Disk, More Flash
[image courtesy of Nexsan]
The E18F is designed and optimised for quad-level cell (QLC) NAND technology. If you’re familiar with the Nexsan E-Series range, you’d be aware of the E18P that preceded this model. This is the QLC Flash version of that.
We spoke about a couple of use cases for the E18F. The first of these was with data lake environments. These are the sort of storage environents with 20 to 30PB installations that are subjected to random workload pressures. The idea of using QLC is to increase the performance without significantly increasing the cost. That doesn’t mean that you can do a like for like swap of HDDs for QLC Flash. Varanasi did, however, suggest that Nexsan had observed a 15x improvement over hard drive installation for around 3-4 times the cost, and he’s expecting that to go down to 2-3 times in the future. There is also the option to use just a bit of QLC Flash with a lot of HDDs to get some performance improvement.
The other use case discussed was the use of QLC in test and dev environments. Users are quite keen, obviously, on getting Flash in their environments at the price of HDDs. This isn’t yet a realistic goal, but it’s more achievable with QLC than it is with something like TLC.
QLC And The Future
We spoke briefly about more widespread adoption of QLC across the range of StorCentric storage products. Varanasi said the use “will eventually expand across the portfolio”, and they were looking at how it might be adopted with the larger E-Series models, as well as with the Assureon and Vexata range. They were treating Unity more cautiously, as the workloads traditionally hosted on that platform were a little more demanding.
Thoughts and Further Reading
The kind of workloads we’re throwing at what were once viewed as “cheap and deep” platforms is slowly changing. Where once it was perhaps acceptable to wait a few days for reporting runs to finish, there’s no room for that kind of performance gap now. So it makes sense that we look to Flash as a way of increasing the performance of the tools we’re using. The problem, however, is that when you work on data sets in the petabyte range, you need a lot of capacity to accommodate that. Flash is getting cheaper, but it’s still not there when compared to traditional spinning disks. QLC is a nice compromise between performance and capacity. There’s a definite performance boost to be had, and the increase in cost isn’t eye watering. StorCentric Announces QLC E-Series 18F
I’m interested to see how this solution performs in the real world, and whether QLC has the expected durability to cope with the workloads that enterprise will throw at it. I’m also looking forward to seeing where else Nexsan decide to use QLC in its portfolio. There’s good story here in terms of density, performance, and energy consumption – one that I’m sure other vendors will also be keen to leverage. For another take on this, check out Mellor’s article here.