Good post by Matt Kixmoeller (thank you) from PureStorage
Over the weekend news broke out of Israel that EMC is potentially close to acquiring XtremIO, as you can read in Globes, Forbes, MassHighTech, BizJournals, ITPro, and earlier The Register. The deal, of course, isn’t official until EMC says so, but in the meantime I’d like to share how we see the emerging market segmentation between all-flash storage and existing spinning disk, and what EMC potentially wading into the all-flash waters says about the evolution of the disk market.
Storage Market Segmentation: Before and After Flash
The storage market has long been divided into “tiers.” Everyone defines them a bit differently, but the relative demarcations in cost per gigabyte and performance are illustrated below. Tier 1 “performance storage” has traditionally been made up of 10K and 15K RPM Fibre Channel or SAS hard drives, more recently front-ended with some flash as a cache or tier. Tier 2 has traditionally meant more capacity-oriented SATA disk arrays. And historically, Tier 0 has been a niche market, served either by ultra-expensive DRAM/flash appliances, or more recently by PCIe flash cards in servers themselves.
As flash has entered the enterprise storage scene, it has been positioned primarily in two ways by vendors:
- Tier 0 flash appliances: all-flash storage devices that are high on performance and cost, and low on scalability and traditional array features (as they are typically used to accelerate a single application).
- Flash caching / tiering as part of a Tier 1 disk array: retrofitting legacy disk-architecture arrays with a small (~5%) flash tier and/or cache to try and improve performance while maintaining economics.
At the lower end of the market, Tier 2, flash is having a strong impact as well. Startups like Nimble Storage are marrying very cheap SATA disks with small flash caches and focusing on ease-of-use and integration to create a storage platform for mid-range customers. This will certainly have an impact on the low-end of the market, but that’s not particularly germane to this discussion.
So What’s Holding Back Flash in the Tier 1 Market?
What you don’t see is many of the existing storage vendors getting aggressive about positioning flash as an alternative to their Tier 1 disk platforms, instead keeping it solidly in the vaunted land of Tier 0, and pitching it as performance, performance, performance. Why is that?
- Flash requires a new array architecture. The existing Tier 1 array vendors have been largely-focused on retrofitting flash into their existing array architectures…but the return on this investment has been modest at best. Arguably the leader in flash-retrofitting, EMC, is now potentially voting with their check book and acquiring a new platform for flash delivery, XtremIO. And the real news is that every other vendor will soon come to the same conclusion. To deliver on the true potential of flash, and to enable flash to go mainstream, a purpose-built platform designed around flash is critical.
- Affordable flash requires high-speed deduplication and compression. Deduplication and compression are technologies that not only reduce data to be able to fit more data on the same amount of flash (reducing usable cost/GB), but also help increase performance in a flash array (by avoiding expensive writes to flash and wear). Although deduplication in backup has been prevalent for years, it is much harder to do inline for primary storage (just ask EMC, they’ve had two leading dedupe technologies, Data Domain and Avamar, for years and have been unable to incorporate them into primary storage). Building an economic Tier 1 flash array requires high-speed data reduction technologies built for flash, something you simply can’t get from legacy Tier 1 array platforms.
- Flash threatens the core disk business. Aside from the technical reasons – there is just as strong of a business reason that legacy array vendors want to keep flash expensive. Existing array vendors source a huge part of their revenues from Tier 1 disk – and they simply can’t sacrifice that for flash. Rather, existing vendors are much better served if they can position flash as a new tier of storage – higher performance, and of course, higher cost than what customers pay today. This both preserves the Tier 1 disk business, and creates an entirely new Tier 0 business to monetize (one that largely competes with server DRAM and server flash).
So in the end, it’s not surprising how the market has played-out to date. Flash has largely stayed expensive as a Tier 0 technology, as the existing vendors lack both the technology and business incentives to push flash aggressively into Tier 1. As a result, the average storage customer has yet to meaningfully participate in the promised “flash revolution,” it has remained a technology reserved for elite applications and elite budgets.
Read on here
