Live from Austin, Texas, Max Mortillaro reports on the FlashArray //C. FlashArray //C joins the Tier 1 storage FlashArray //X platform, and the high-performance FlashBlade object & file storage platform.
FlashArray //C is Pure Storage’s response to the challenges with hybrid secondary storage platforms, namely management complexity, inconsistent performance and a lack of enterprise features (all according to Pure).
Pure Storage claims that //C was born out of listening to customer demand, namely a need for an end-to-end Pure Storage experience across platforms and storage tiers.
FlashArray //C in a nutshell
At the time of release, we haven’t been given much technical details (a deep dive will hopefully follow later on) but FlashArray //C is essentially using QLC 3D NAND memory over NVMe, and uses 3D XPoint for caching.
When purely look at the flash media itself, QLC NAND offers the highest density, but is also considerably slower than TLC, and it also suffers from a lower endurance. These downsides are mitigated by a lower $/GB cost compared to TLC, and let’s also not forget that the array also uses Tier-0 3D XPoint flash to give a performance boost.
Don’t get mistaken though, //C is a secondary storage platform, where performance characteristics are less critical than with primary storage. Pure claims a “predictable 2 to 4ms latency”, and a 10x consolidation ratio over hybrid / HDD based platforms.
The platform currently offers three capacities: 1.3 PB, 3.2 PB and 5.2 PB, all rated as effective capacities (i.e. after applying data reduction mechanisms; raw capacities are 366 TB, 878 TB and 1.39 PB).
The QLC flash capacity is provided through Pure Storage proprietary DirectFlash form factor. It is not clear at this point how the Optane (3D XPoint) capacity is provisioned.
We also haven’t been given any deep dive as to how Purity will handle write operations to reduce wear on the low-endurance QLC flash (with less than 1K P/E cycles), but it’s expected that they would use common mechanisms they’ve used in their other arrays such as a global FTL (Flash Translation Layer) and other data reduction techniques.
The FlashArray //C is targeted at classical secondary storage use cases. Pure Storage advocates for FlashArray //C use in the context of Tier-2 applications Disaster Recovery, Policy-based VM tiering, Multi-Cloud Dev/Test, snapshot consolidation, and use as a backup target.
FlashArray //C can be consumed as block (FC, iSCSI, NVMe-oF) and file storage (NFS, SMB). It also supports vVols, a VMware technology that Pure Storage has been one of the greatest proponents.
Currently, all we know is that the FlashArray //C will use the Pure1 analytics platform. It remains yet to be seen whether this platform will continue to exist “as is”, or if Pure will consider baking in some data management capabilities.
The next hypothesis here is (assuming they’re considering that) what would be the best way to achieve this. I cannot say if Pure Storage have enough in-house resources to work on such a project, therefore it is likely that if they envisioned this route, they might need to achieve these capabilities through an acquisition.
I’m interested in understanding what Pure Storage will be working on at the software level. Data management platforms are all the rage, and going into this direction may help Pure Storage expand laterally into this lucrative market.
It’s interesting to see how technology can be used to achieve different outcomes. We’ve recently seen startup companies advocate the use of 3D XPoint and QLC 3D NAND to build a massively scalable and fast single tier storage.
Pure Storage instead took the decision to build a secondary storage platform while using the same components. It’s understandable that building a high performance solution based on 3D XPoint and QLC would be a way of shooting themselves in the foot, considering the existence of the //X and of FlashBlade.
There are already a lot of secondary platforms out there, but not much that I know of using QLC and 3D XPoint. The footprint gains are real, the necessity to replace HDDs with QLC is another topic. Are HDDs doomed? Perhaps not, but it seems foreseeable that they will be displaced to a lower tier (archival, perhaps) by QLC.
By the way, I’m curious to hear if Pure is going to use Optane in the //X or in FlashBlade. Looking at our badges, Optane is mentioned so looking forward to the other announcements.
FlashBlade brings additional capabilities to Pure Storage’s product line and allows them to deliver a consistent, end-to-end experience to their customers.
The next step will be to understand what capabilities this product will have, i.e. if Pure Storage will have the boldness of baking in data management capabilities. Stay tuned!