Earlier this week EMC announced a coming update to XtremIO, slated to ship sometime in the 2H of 2014. This is one of a few recent competitor announcements that are showing the evolving maturity of the all-flash array space. For our summer update on the market at large, see our Market Evolution post, also issued today.
In that post we laid-out a 10-item “recipe” on what we believe constitutes a market-competitive all-flash array in 2014. So does the 3.0 release bring XtremIO up to the level of a true All-Flash Array? It is hard for us to comment specifically since the release hasn’t shipped and details are light…we’ll see when it makes its way into customers hands as GA. In the mean time, we’ll arm you with some key questions to ask EMC about XtremIO 3.0 yourself, so you can assess XtremIO’s readiness for your production workloads:
- Dual-drive failure recovery. This was announced as part of the XtremIO GA in 2013, has it shipped yet? What happens if two drives fail in an xBrick simultaneously?
- Online capacity expansion. If my storage footprint grows, can I expand storage by adding Xbricks non-disruptively, or do I have to migrate off all my data, re-initialize a larger cluster, and move my data back? If I start with 10TB xBricks, can I add 20TB xBricks or future larger bricks to scale as my data grows?
- Maximum deduplication rate / virtual size. What is the maximum virtual user data size of a 10TB and 20TB xBrick? Is there a limit to the maximum data reduction I can achieve when full? Can the 20TB xBrick hold twice the amount of user data as the 10TB xBrick at maximum data reduction when full? Do snapshots count against this, and thus is there a limit to the number of snaps?
- Compression enablement. The new release will come with an upgrade to compression for current customers. Can I enable it non-disruptively, or do I have to migrate all my data off and start over?
- Encryption enablement. The new release will come with an upgrade to encryption for current customers. Can I enable it non-disruptively, or do I have to migrate all my data off and start over?
- 4K vs. 8K block size. The performance demonstrations shown during EMC’s launch event were all driven with 8K IO vs. XtremIO’s historical 4K-optimized architecture. This leads us to believe that XtremIO changed its native block size from 4K to 8K to enable compression in XIOS 3.0. If true, how does this impact scale and performance across IO sizes?
- Dedupe effectiveness. If the native block size of XtremIO did indeed change from 4K to 8K, how does this impact the effectiveness of deduplication? Was deduplication quality and effectiveness sacrificed to enable compression? What level of reduced deduplication effectiveness should I expect at the potentially larger 8K fixed chunk size?
- Performance improvements. EMC made a big deal around performance improvements in the XtremIO 3.0 release. Were there fundamental improvements that delivered this, or was it just the potential change from 4K to 8K metadata (requiring less metadata operations for a given amount of data)?
- Snapshot metadata garbage collection. In previous EMC tech talks, XtremIO’s CTO discussed the overheads of XtremIO snapshot garbage collection, where deleting snapshots could take hours to garbage collect excess metadata. Has this been fixed in XIOS 3.0?
- Snapshot policy automation and deletion. How are snapshots scheduled, managed, and deleted?
- Extreme performance loss through failure. EMC videos of HA demonstrations during the XIOS 2.x era showed a 50-75% performance loss on HA events (despite an articulated design for much lower). Has this been improved in 3.0?
- VPLEX / RecoverPoint Tax. Since EMC has chosen not to ship native replication for XtremIO, what is the cost burden of the VPLEX and RecoverPoint appliances and software licenses required to deploy into a DR scenario? Do I need a VPLEX engine for EACH XtremIO xBrick to maintain full performance? What’s the performance impact of putting all that gear in front of XtremIO? What’s the cost impact? Does RecoverPoint take advantage of the data reduction capabilities of XtremIO?
Ask EMC these questions yourself, and better yet, test them during your PoC! Fundamentally, the largest question you should ask yourself is: Are you comfortable with the level of disruption that seems to come with every new XtremIO feature? Pure Storage didn’t GA our product until we had non-disruptive upgrades of everything, it just isn’t viable to run a storage array in production where upgrades require downtime, let alone data migration.
That’s it for now…enjoy the rapidly-maturing all-flash array market, it’s never been a better time to be a storage customer, and it’s never been a better time to try flash!
Note: The Pure Storage and “P” logo marks, Purity Operating Environment, and RAID-3D are trademarks of Pure Storage, Inc. All other trademarks or names referenced in this document are the property of their respective owners.