Demand-Driven Storage Blog

Dynamic or automated storage tiering – it seems like everyone’s talking about it, but no two vendors describe it the same way. Over the past few weeks, we’ve been busy briefing analysts and press about the new Avere Systems FXT Series. This group tends to be a skeptical lot by nature and most don’t have the ability to implement new products in order to put vendor claims to the test like prospective customers can. So when we started talking about the performance and cost-savings benefits of moving data between fast media tiers in our storage appliance and slower media in a traditional NAS filer, many held up their hands and asked us to define exactly how Avere’s tiering was the same or different than: Compellent’s data progression; or what EMC announced with FAST, etc. Based upon these discussions, we came up with four questions to ask any vendor (including Avere) who offers dynamic or automatic storage tiering:

1. How much human intervention is required in order to move data between tiers? Does the system require someone to tell it every time they want data to move from one tier to another? Are policy configurations required upfront – i.e. an administrator must map applications or other data attributes to tiers and save those settings within a system in order for data movement to occur. In this case, policies stay in effect until they are modified or disabled by the administrator. Or is it truly automatic, meaning policies that take into account frequency of access, file size and other characteristics are built into the system as a form of intelligence and require no action on the part of an administrator?

2. How rapidly is data moved from one tier to another? Are we talking weeks, days, hours or on-the-fly? The answer to this question can have a dramatic impact on any performance benefits you may hope to gain from dynamic tiering. Of course, the answer to the question above will help give you hints as to how or if a solution will have the ability to perform real-time data movement.

3. At what granularity can data be moved from one tier to another? Are we talking LUNs, volumes, files or blocks? The finer the granularity of data that can be moved onto faster tiers such as Flash, the more dramatic the benefits will be in terms of both application performance and cost savings. If a solution is limited to moving LUNs around, it obviously wouldn’t be very economical to add SSDs to the array since you’d be overpaying to promote all blocks in the LUN regardless of the frequency of access.

4. And finally, can data be moved to tiers outside of a specific vendor’s solution? If storage tiering only works inside of that vendor’s system and won’t play well with others, that’s great news for that vendor because they’ve got you locked in, but not so great news if you’re looking for the potential benefits that interoperability often provides.

Did I miss anything – are there more question you’d add to the list?

BTW – if you’re curious about how Avere’s dynamic tiering works, check out this product brief written by Terri McClure of ESG.