EMC Announces XtremIO 3.0 - Architecting IT

This week EMC held a launch event in London branded “Redefine Possible” covering announcements on VMAX, Islion, XtremIO and some of the software platforms. This post covers a discussion of the changes to XtremIO.

Background

XtremIO was a startup company acquired by EMC in 2012, with the first products reaching general availability on 14 November 2013. In some respects the architecture of the product is similar to other EMC platforms such as VNX and VMAX, comprising a pair of active-active controllers, battery backup and DAE (Disk Array Enclosure), combined to create what the company calls an “X-Brick”. Internally there are significant architectural differences, including a redesigned RAID-like data protection mechanism called XDP and a content addressable data storage model. Version 2.4 of the XIOS operating system (which runs XtremIO arrays) was announced at EMC World 2014 and was generally available as of the EMC event in London. This update included support for snapshots and data-at-rest encryption through SED (Self Encrypting Drive) SSDs.

Version 3.0

The London announcement was all about features in the 3.0 release of XIOS, expected to be generally available later this year. These features include:

Maximum capacity of 120TB raw, through increased scale to six 20TB X-Bricks.
New “baby” X-brick using 13 400GB flash drives for 5TB capacity, upgradable to 10TB.
Data Compression – inline compression giving 2x – 4x savings, depending on data type.
Writeable Snapshots.

The XIOS code has been tweaked and EMC now claim 2x SSD rebuild, 2x bandwidth (but not reduced latency), 2x faster VM provisioning and 50% more database performance than on the previous code base.

If we look at how these features compare, we see other platforms with equal or higher scalability (Kaminario K2, HDS HUS VM, HP 3PAR 7450), platforms that already provide compression and de-duplication (Pure Storage FlashArray, Nimbus Gemini, SolidFire, Kaminario K2) and platforms that offer writable snapshots (HP 3PAR 7450). Data compression isn’t a killer feature either; most arrays offer de-duplication, compression or both, at similar savings levels. This makes the 3.0 upgrade a catch-up rather than leapfrog release. One final point; we still saw no mention of the ability to field upgrade and add nodes to a running XtremIO cluster.

Number 1 – Already?

What was probably most confusing was EMC’s claim that XtremIO is both the number 1 all-flash array in the marketplace, plus the fastest growing storage array in terms of revenue of “all time”. When I challenged Jeremy Burton on the follow up CrowdChat, he indicated that there were no official figures available to make this statement and the premise was made based on insider knowledge of competitor’s sales. Bearing in mind at EMC World, EMC claimed to be the leader in flash shipments only for HDS and IBM to demonstrate better numbers, then you can make your own assumptions on the XtremIO claims.

Guarantees

As part of the announcement EMC “extended” the $1m XtremIO Flash Guarantee announced at EMC World to the newly announced features. As we know this challenge is nothing more than a marketing stunt, so let’s move on. What’s more interesting are EMC’s latest set of guarantees:

Seven year maintenance price protection
Three year money back guarantee
Seven year flash endurance program

All of these offerings are designed to fight back at the competition and it’s clear EMC realise they can’t continue with previous practices, especially around maintenance hikes.

The Architect’s View

EMC are making solid progress in adding features to XtremIO and as the feature gap narrows compared to the competition, other factors will no doubt come into play, such as usability and integration into existing environments. With the additional announcements around VMAX³, the gap between the features offered by XtremIO and the new VMAX seems to have closed up again. The only difference I can see here is in the area of latency; presumably XtremIO delivers I/O at a lower latency that can ever be achieved with VMAX. However I wonder if latency alone is enough to justify an entire separate platform. Certainly for customers the choice of VMAX or XtremIO is less clear than it was a few days ago.

Background

Version 3.0

Number 1 – Already?

Guarantees

The Architect’s View

Related Links