EMC goes all in with DSSD

by Robin Harris on Monday, 5 May, 2014

The drama with XtremIO’s delays only whetted EMC’s appetite for more drama. Or the hot breath of competition and declining VMAX sales.

Let’s go with the latter.

StorageMojo has been following DSSD for some time. The core of the Sun ZFS team, Jeff Bonwick and Bill Moore, are founders, and Andy Bechtolscheim an investor and advisor.

In the post on DSSD, StorageMojo opined:

So what are they building? They are taking a radically different approach to the problem of high-performance transaction processing storage. The use of flash is a given in TP, and the extra durability, scalability and guaranteed read latency would be very attractive in large TP applications.

The EMC presser:

Menlo Park-based DSSD is the developer of an innovative new rack-scale flash storage architecture for I/O-intensive in-memory databases and Big Data workloads like SAP HANA and Hadoop.

EMC and StorageMojo agree! Will it last? No.

EMC also said:

Menlo Park-based DSSD will operate as a standalone unit within EMC’s Emerging Technology Products Division reporting to Chirantan “C.J.” Desai. DSSD President and CEO Bill Moore, formerly Sun Microsystems’ Chief Storage Engineer, ZFS co-lead and 3Par’s first employee, will lead the DSSD business within EMC. . . .

Andy Bechtolsheim said, “The prospects of what EMC and DSSD can achieve together are truly remarkable. We ventured out to create a new storage tier for transactional and Big Data applications that have the highest performance I/O requirements. Working together with EMC, DSSD will deliver a new type of storage system with game-changing latency, IOPS and bandwidth characteristics while offering the operational efficiency of shared storage.”

Good, they’re calling it what it is, an emerging technology. Nor are they being as aggressive on the timeline as they were with XtremIO:

Products based on the new DSSD rack-scale flash storage architecture are expected to be available in 2015 and will be optimized for:

    In-memory databases (e.g. SAP HANA, GemFire, etc.)
    Real-time analytics (e.g. risk management, fraud detection, high-frequency applications, Pivotal HD, etc.)
    High-performance applications used by research and government agencies (e.g. genomics, facial recognition, climate analysis, etc.)

Customers desiring a platform capable of delivering unprecedented performance for I/O- intensive Big Data and in-memory applications like SAP HANA and Hadoop will choose DSSD rack-scale flash storage as the fastest tier in their multi-tier storage architecture.

The StorageMojo take
Joe Tucci is reinventing EMC one more time. Will it work?

XtremIO is the VNX replacement. DSSD is the VMAX replacement.

It’s been clear for years that standard RAID arrays – say, VNX and VMAX – are on the way out. Object/cloud storage is winning the bulk storage market and flash-based storage the high-performance market.

This is a return to the EMC of the mid-90s, where a young and flaky Symm was positioned as the high-performance alternative to IBM’s tired mirrored disks. As soon as others caught up on performance, EMC promptly repositioned it.

DSSD’s hardware development has not been trouble-free, despite the promising architecture, patents and dev team. Sometimes you just need to lay out a PC board, and if high performance is the goal that takes serious and specialized experience and skill.

Competitors and would-be competitors need to rev up their engines. You’ve got a year to get a footprint in as many EMC customers as possible before they bring this new VMAX hammer down.

Courteous comments welcome, of course. Andy’s done a masterful job at winning graceful exits before. I expect this is another.

{ 3 comments… read them below or add one }

Simon Sharwood May 5, 2014 at 4:18 pm

The signs have been there for a while that EMC wants a VMAX successor – last year VMAX was described – http://www.theregister.co.uk/2013/08/13/emc_demotes_vmax_into_capacity_tier/ – to me as a mere “capacity tier”. XtremIO is powerful, but not VMAX-replacement powerful. I’m a bit grumpy with myself for not spotting that sooner 😉

Dave Wright May 5, 2014 at 6:03 pm

I think you’re right that EMC needs a VMAX replacement, but I don’t think this is it.
VMAX is really about RAS and feature set, not just performance.
While DSSD will probably have better performance than XtremIO, expect it to have far less feature-set wise and be highly tailored for specific vertical applications rather than general purpose storage.
This is the performance-at-any-cost tier, versus the “performance and capacity” tier of XtremIO and the “capacity only” tiers of Isilon and Atmos.

Brett May 15, 2014 at 2:23 pm

From the rumblings I’ve heard, EMC will have a VMAX “3” out by year end. So I doubt this will be a VMAX replacement.. Remember, the VMAX is not a performance-centric array. Though it’s no slouch, it is not really sold as such. It is sold as a do-everything (think: Swiss Army Knife), support-everything, always-on massive slab of storage. If you are only looking for speed, there are many many other options out there. Will DSSD attach to mainframe, AS400, and open systems? I doubt it.

Also, how can XtremIO be a replacement for the VNX when it doesn’t offer any file services? Perhaps, one day they will stick a VNX Gateway in front of it but that defeats some of the scale out functionality of the array.

My guess is that they will not release a DSSD “array” at all. They will use the IP developed in future revs of existing products.

Leave a Comment

Previous post:

Next post: