Thursday, June 27, 2013

What is Dynamic Disk Pooling?

When IBM introduced firmware release 7.8x in the fall of 2012, there were a number of new technical advances, but by far the most exciting was Dynamic Disk Pooling.  With the general release of firmware level 7.84, many of these capabilities became available not only to new IBM System Storage DS3500 purchases, but to installed systems as well.  DDP does not require a license, nor does it require a purchase.  As soon as a DS3500 system is upgraded to 7.84 or later (7.86 is the latest), Dynamic Disk Pooling is available.

DDP dynamically distributes all data, spare capacity, and protection information across a pool of drives.  That pool may be all drives in the system, or a subset, such as all of the SAS 15K drives.   Effectively, DDP is a new type of RAID level, built on RAID 6.  It uses an intelligent algorithm to define where each chunk of data should reside.  In traditional RAID, drives are organized into arrays, and logical drives are written across stripes on the physical drives in the array.  Hot spares contain no data until a drive fails, leaving that spare capacity stranded and without a purpose.   In the event of a drive failure, the data is recreated on the hot spare, significantly impacting the performance of all drives in the array during the rebuild process. 

With DDP, each logical drive’s data and spare capacity is distributed across all drives in the pool, so all drives contribute to the aggregate IO of the logical drive, and the spare capacity is available to all logical drives.  In the event of a physical drive failure, data is reconstructed throughout the disk pool.  Basically, the data that had previously resided on the failed drive is redistributed across all drives in the pool.  Recovery from a failed drive may be up to ten times faster than a rebuild in a traditional RAID set, and the performance degradation is much less during the rebuild.

Apart from improved data protection, performance, and failure recovery, there are other benefits from DDP as well.   Administration is significantly easier.  When adding capacity, there is no more agonizing over whether to create a new RAID set or expand a current RAID set, and no more question about where the optimum placement is. You just put in the drive and add it to a pool.  The total capacity of the pool has now been increased, and you can then proceed to expand your logical drive, or create a new logical drive as needed.  Using disk pools instead of RAID sets makes disk utilization much more efficient and avoids issues found with traditional RAID, such as islands of stranded capacity that can’t be recovered.

With the proliferation of large format drives, traditional RAID is becoming progressively more difficult to manage, and the long rebuild times associated with large SATA drives creates unacceptably long windows of vulnerability (as much as 4.5 days with a 3TB drive in an operational array).  Over the coming months and years, technology like DDP will continue to replace traditional RAID because of these issues.

No comments:

Post a Comment