Auto tiering is used to help optimize data running across LUNs and file systems. It’s an important storage efficiency tool, and many users and vendors see it as a must-have technology. It’s especially key for users who have solid-state drives (SSDs) in place. SSDs have capacity constraints, and auto-tiering software helps with performance demands.
In this podcast interview, Jeff Boles, senior analyst and director, validation services at Taneja Group, discusses the importance of auto tiering in today’s data storage industry. Find out auto tiering’s main purpose in the data center; the difference between sub-LUN tiering and auto tiering; when sub-LUN tiering is a better choice than auto tiering; who the big vendors are that support auto tiering; and what the future looks like for automated storage tiering. Read the transcript below or download the MP3.
SearchStorage.com: What is the main purpose of auto tiering in the data center?
Boles: Auto tiering is the automated ongoing optimization of data across LUNs or, conceivably, file systems. Basically, the hottest data blocks are tiered to higher performance media, while the coldest data is [sent to] lower cost, higher capacity drives.
SearchStorage.com: What’s the difference between sub-LUN tiering and auto tiering?
Boles: They’re very similar. I consider sub-LUN tiering a sub-category of auto tiering. Both work to optimize data across different tiers of disk. Sub-LUN works with volume blocks or segments (called various things depending on the vendor) within an individual volume. Auto tiering typically applies to entire volumes or entire LUNs.
SearchStorage.com: When is sub-LUN tiering a better choice than regular auto tiering?
Boles: Sub-LUN tiering can optimize the performance of many different LUNs with a limited pool of high-performance resources. In contrast, auto tiering moves entire volumes or much less granular sets of data; it can quickly consume what might be all of a limited resource such as SSD. With array-based SSD costing what it does today, this might mean auto tiering doesn’t meet your needs sometimes. But let me clarify that just a bit. This is a bit of [a fuzzy] line for some vendors -- this message about sub-LUN being more efficient and more granular and helping you to do more with less. But with some systems, SSD is cost-effective enough and can be scaled to large enough quantities that whole-volume auto tiering isn’t necessarily a bad thing. Moreover, it can reduce the amount of overhead and data scattering that happens across your volumes. This can reduce complexity and clear up just what a drive failure means to you, and make things a little easier to manage. So it’s not that either one is necessarily good or necessarily bad; it depends on what you’re trying to accomplish. But sub-LUN tiering is more granular. It can help you make better use of a pool of SSD by leveraging across many different volumes a bit more efficiently than whole-volume auto tiering will do.
SearchStorage.com: Who are the prominent vendors in the auto-tiering market?
Boles: Auto tiering is pretty much becoming a must-have technology with everyone having some form of it. EMC Corp. is certainly leading the market by introducing the same core functionality and a broad range of products. IBM is another leader in this space. And Dell has done some really great stuff in its EqualLogic arrays. NetApp has a really interesting approach that leverages SSDs as an acceleration tier in a unique way by migrating up hot blocks that are accessed and then using SSD as a rig cache. This can yield good results.
But every major vendor out there, even smaller ones, has [auto-tiering] functionality. And any list that hopes to be comprehensive on that is bound to leave some out. So expect [auto tiering] from your vendor, and look for everyone to have it eventually.
SearchStorage.com: What does the future look like for auto tiering in the storage industry?
Boles: With the direction we’re going, it’s soon going to be a core functionality in every competitive storage array. I think the real leaders and innovators in this space simultaneously realized that what it takes to do smart auto tiering gives them a unique opportunity to deliver really innovative I/O analytics and better I/O management to boot. This can make storage a much more efficiently used resource, which is something that industries are clamoring for and where auto tiering has gained its traction. Instead of storage blocks always stepping on each other and you [not being able to] differentiate between one workload and the other, auto tiering can finally make storage a more efficient shared resource, much like virtualization has done for the server. It helps you delineate between workloads and optimize them according to their demand. That’s better than what we were ever capable of before.