Thursday, April 15, 2010

Automated tiering leverages SSD to deliver high end performance and typically, high capacity SATA to deliver economy. While there are a finite number of data centers that can take advantage of a broad based application of SSD to improve performance, most have inactive data that can utilize high capacity SATA to drive down overall storage costs. The question is, "Is leveraging SATA in the primary storage system the best way to achieve that goal?"

The advantage the automated tiering storage systems bring is the ability to move this data between storage tiers without administrator intervention and to do so transparent to the user and without interruption. In an environment where time to monitor data use trends and time to act on those trends is limited, this can be an appealing option. There is a temptation to use automated tiering as the sole mechanism for storing aging data, instead of leveraging disk archiving systems that were purpose-built to do so. However, this approach may have some issues.

Automated tiering provided from the storage system manufacturer is the ultimate in 'vendor lock-in'. For the automated tiering to work, all the disk, be it SSD, fibre or SATA drives, typically must come from that vendor. While storage system vendors do provide their SATA drives at a reduced cost, when compared to their fibre drives, they're not often as price competitive as vendors who provide systems that were designed for the archive tier. 

Disk archiving or value-tier focused vendors achieve their pricing delta in a number of ways. While most vendors use the same SATA drives, these vendors can afford to price these drives at a lower cost because they don’t have the same overhead as larger vendors do and they offer a high level of support and reliability. Also, these vendors systems that the drives are designed to be housed in are less expensive than the drive shelves that primary storage vendors use. This is a case where design meets function, and a more complex design isn't needed. SATA drives, which in most cases are the sole drive type in the archive tier, don't have the same performance capabilities of SSD or fibre drives. Most primary storage suppliers deliver a drive shelf that can handle SSD, fibre or SATA in the same shelf and in many cases mix and match those differing technologies. There are additional engineering costs in designing drive shelves that are flexible enough to handle these drive combinations. Disk archiving solution vendors, on the other hand, can focus on a simpler and sometimes more elegant approach via a single technology, thereby reducing costs.

Beyond hard cost advantages, disk archiving vendors also have the capability to optimize storage when it's first written. For example, Permabit Technology, with its Enterprise Archive products, compresses and deduplicates data inline and optimizes the data storage costs, footprint, power and resource utilization. By applying these storage optimization techniques, data being stored on archive storage can be reduced by as much as 5X and in some use cases greater than 20X, depending on the data type. Optimization creates an even greater cost advantage for the archive tier. 

Also, related to this cost advantage is the ability of a purpose built archive system to scale to very high overall capacities. While many primary storage systems can scale to the low hundreds of terabytes, archive systems are designed to scale to the thousands of terabytes. This makes the above cost advantage even more important. If the capacity of the archive tier is half the cost of the SATA capacity on the primary tier, then the savings at 50TBs is impressive, the savings at 500TBs is critical. Particularly in today’s budget constrained IT departments.

In addition, there is the matter of the function these systems are required to perform. A key capability of archive systems is the ability to perform periodic consistency checks of data on a regular basis; many primary storage systems do not do this. As a result, if data that has been idle for five years and is suddenly needed, there is no guarantee that the data will be accessible or even that the data is the same as it was originally written. Archive systems do perform this consistency check and if degrading drives are found, data is seamlessly and automatically shifted to safer drives. In addition, if there are any compliance demands where data has to be retained for a certain number of years and/or that data needs to be stored in a manner that it can't be altered, most primary storage platforms do not have that capability where archive platforms do! 

As stated earlier, a key advantage that primary storage systems with automated tiering have is the tight integration of data movement. But, this comes at some loss in control and understanding about where the data actually is residing. While a broad application of tiering can be beneficial at times, there are other times when knowing exactly where a data set is, when it was moved and how long it has been there are critical. This can all be driven, and even automated, when disk archiving solutions are coupled with software applications that perform email, file or database archiving. These software applications are all designed to work with archive-specific storage.

Finally, for other types of data where a broad based application of data movement is more desirable, that capability can be integrated with a traditional archiving system via the use of a secondary appliance that will move data between the primary storage and archive storage. This allows both systems to be used for what they are best at and may actually provide the greatest balance of performance, data protection and cost savings.

Automated tiering has its place in the data center when accelerating performance through the simpler integration of SSD drives and to slightly lower the costs of storing near-active data. For data centers facing large amounts of inactive data that must be stored securely long-term, purpose-built disk archiving solutions deliver the best performance and are the most cost-effective.

George Crump, Senior Analyst