Automated Tiering: Cost Savings Or Performance Enhancer?

One of the questions that came up quite a bit in response to my last entry was whether or not Automated Tiering Systems (ATS) were designed to drive cost savings or to be a performance enhancers. In that entry, I took the stance that ATS is a performance enhancer, while some of the comments and emails suggested that it was a cost enhancer. The real answer is that it's what you make of it.

George Crump

March 2, 2010

3 Min Read
Network Computing logo

One of the questions that came up quite a bit in response to my last entry was whether or not Automated Tiering Systems (ATS) were designed to drive cost savings or to be a performance enhancers. In that entry, I took the stance that ATS is a performance enhancer, while some of the comments and emails suggested that it was a cost enhancer. The real answer is that it's what you make of it.

As I have written in the past, SSDs themselves can even provide a cost savings versus 300 short-stroked 15k RPM disk drives. Since ATS moves data up and down storage tiers based on activity, it should allow you to purchase less SSD and to make sure that SSD has only the most active data. It should also allow you to purchase significantly less Fibre Channel or SAS-based drives in favor of SATA, another cost savings.

When ATS capabilities are fully built-out from the manufacturers, they should be able to deliver both high performance AND cost savings. For this to work, though, there are two missing components in some of the manufacturers' solutions. The first is a level of granularity. Moving an entire LUN up and down storage tiers is simply not going to work. Data movement has to be done at a granular level. Second, there has to be some sort of intelligence behind what gets moved. You can't move bits of files to SSD because they are being accessed. One or two accesses does not in most cases justify promotion to SSD. What if they are word documents being accessed by users on WiFi connected laptops? Unless thousands of those users are accessing the same file at the same time, they will never see benefits from the SSD performance. The system also needs the ability to automatically decide what specific active files should be promoted to SSD, with some manual overrides. This applies with demotion to SATA as well. There may be some files that, while infrequently accessed, make more sense to have on Fibre or SAS. Again, the system has to provide either automation or insight to help the users make those determinations.

The exception is if you have excess capacity on your SSD. You want to use ALL of your SSD capacity to justify the expenditure. Free space on SSD really is wasting money. The system also has to be intelligent enough to be more flexible when it has plenty of SSD capacity at its disposal. That may not be the case with ATS installed this year, but in years to come it very well may be as SSD prices continue to drop.

The other interesting aspect of ATS is how the various vendors are delivering the capability. There is the cache-like functionality that some argue is not automated tiering, then there is the more permanent technique used by some of the storage vendors. As always seems to be the case, in storage there is a disagreement on what the term really means. In our next entry we will go through some of the methods by which vendors are claiming to deliver automated tiering.

Read more about:

2010

About the Author(s)

SUBSCRIBE TO OUR NEWSLETTER
Stay informed! Sign up to get expert advice and insight delivered direct to your inbox
More Insights