Mark Adams

What a Buzz Kill - Xtremly Disappointing!

Blog Post created by Mark Adams on Sep 3, 2015

There’s been a tremendous amount of buzz about flash storage this year and all of the talk coming out of VMworld and the Flash Memory Summit has only added to it.  Many believe that enterprise storage is on the cusp of a change as big or bigger than the transition from direct attached storage to SANs.

 

EMC has been very actively engaged in this discussion and their president of global systems engineering, Chad Sakac, posted an update about XtremIO from VMworld 2015 on his blog—Virtual Geek.  What really caught my eye was this.

 

     VG.png

VG2.png

VMworld 2015: XtremIO - what's new? - Virtual Geek

 

Did he really write this?!?!  If your sales strategy is to lead with Xtrem, shouldn’t your engineering team harden the product first before deploying to thousands of customers?  Are they asking their customers to become part of their QA processes?

 

In reality, Chad is only confirming the evidence we’ve seen that XtremIO isn’t ready for prime-time. Here is a summary:

 

  • Scaling out from a single to multiple x-bricks is disruptive and requires that all data be migrated off and then back.

XIO4.png

https://community.emc.com/message/899508#899508

 

  • Similarly, system upgrades to XtremIO 3.0 were destructive and data needed to backed up and then restored.  EMC received a lot of bad press about this and promised to never again put their customers through such a botched upgrade process—however, see point #1.

ElReg.png

No biggie: EMC's XtremIO firmware upgrade 'will wipe data' • The Register

 

 

  • A single X-brick could not tolerate a dual drive failure without incurring data loss. This risk was finally addressed but not until 4.0 was released.  Part of the hardening process I guess.  Here’s their explanation where they consider a dual drive failure to be a “corner case”:

XDP.png

 

https://community.emc.com/message/863150#863150

 

  • If one Xbrick in a cluster goes down, then the entire cluster becomes unavailable since the data is spread across the cluster.  So much for high availability.

 

Avi.pnganswer.png

https://community.emc.com/thread/206964

 

Here’s what we say.

 

xtrem.png

HDS has shipped well over 200PB (petabytes) of flash capacity because we have customers from whom we’ve earned their trust to provide them with enterprise ready flash storage. We have customers buying multiple petabytes every quarter and are either on the path to an all flash datacenter or already there.  Our customers know that they can run their business on Hitachi flash storage.  The reason they do this is that our Hitachi Accelerated Flash is not only fast (and predictably so) it’s also always available.

Outcomes