Pure Storage Natively Unifies Block and File Protocols on FlashArray

Pure Storage Natively Unifies Block and File Protocols on FlashArray

Chris EvansAll-Flash Storage, Data Practice: Data Storage, Enterprise, Pure Storage, Storage, Storage Hardware

This week, Pure Storage announced an update to FlashArray that delivers native file and block protocol integration.  What benefits does this enhancement offer in a world that has lived with multi-protocol support for many years?

Background

The early days of purpose-built storage appliances (also known as Integrated Cache Disk Arrays) were focused exclusively on block protocols.  Until the widespread adoption of Fibre Channel and storage area networks, shared storage connectivity used SCSI and point-to-point cabling.

Auspex, followed quickly by NetApp, introduced network-attached storage appliances, and it wasn’t long before storage networking across both Ethernet and Fibre Channel was widely available.  However, some vendors, such as EMC (now Dell), implemented multi-protocol through an additional appliance (or software stack) sitting on top of a block-based array.  This solution effectively carved off a pool of storage for file-based usage. 

NetApp implemented block protocols on file systems (an inverse of the transition seen elsewhere).  Across the industry, file and block weren’t generally natively integrated, partially because the I/O profiles of block and file access were subtly different.  In a world of disk drives, a “noisy sequential I/O neighbour” would be an effort to manage.

Workloads

Fast forward to 2023, and as we all know, all-flash systems are where the market is inexorably headed.  Almost all traditional workloads are virtualised, mainly on VMware vSphere/ESXi.  VMware supports block, file, and vVol-based storage solutions (vVol being a managed block solution).  There are benefits and disadvantages in using either block or file as a VMware datastore; however, IT organisations generally choose one or the other based on operational preferences. 

FlashArray

In the latest software upgrade, Pure Storage now supports file and block (including vVols) natively.  What does “natively” mean in this sense?  It means no partitioning of physical space into file and block pools.  Capacity is available to either protocol and not fragmented or wasted.  It also means being able to apply policy-based management to folders in the same way vVols are managed, thereby implementing VM-level policies to applications.  Customers now have the flexibility to use a protocol of choice while retaining the benefits of a single platform, even if those choices change over time.

It’s worth also mentioning that consolidated file and block protocol support isn’t just about server virtualisation. Many businesses need general file storage and don’t want multiple platforms. FlashBlade is designed for large capacity and scale, so having the capability to offer general file with block in FlashArray is a sensible approach (and has been in place for many years since Pure Storage acquired CompuVerde).

The Architect’s View®

Native support of file and block protocols on a single platform might not look like a big deal.  There’s a clear expense saving that can be achieved, due in part to a reduction in fragmentation, but also the ability to support any VMware storage protocols (and not maintain separate platforms).

As with many recent enhancements to FlashArray and FlashBlade, Pure Storage continues to update both platforms and fill the gaps in support and compatibility that remove any barriers in the sales process.  These changes increase the “TAM” of customers, making both solutions more competitive and enabling Pure Storage to sweep the floor of competitive hardware as part of a storage refresh.  This strategy isn’t about technology but about good business sense.

What’s next?  We talked about databases being the 4th protocol in a recent blog post.  Conceivably, we could see Portworx Data Services and PX-Enterprise being more closely integrated with FlashArray.  Imagine rather than using block storage via CSI to provision container storage, an administrator could opt for a file-based option.  In either case, the attributes assigned to a container for storage performance could be aligned with policies at the hardware level, much as vVols work today.


Copyright (c) 2007-2023 – Post #a3bd – Brookend Ltd, first published on https://www.architecting.it/blog, do not reproduce without permission.