One of the problems with modern disk arrays that support thin provisioning is that volumes tend to get fat over time, just like us Americans! Instead of putting yourself on a diet, you can put your storage array on a diet. Think you will need to haul your array to the gym and shove it on the treadmill? Nope! 3PAR announced today a set of technologies that help keep thin provisioned volumes thin, even as data is written and deleted over time.
In a traditional thin provisioned enabled array if you have a 2TB LUN and write 1TB of data, then delete 800GB, that 800GB is still allocated to that volume…stranded. Over time that LUN can get fatter, depending on how data is written and deleted. If you have snapshots of those volumes, they can get fat too. Fat everywhere!
You can check out the announcement at 3PAR regarding their thin conversion, thin persistence, and thin copy reclimation. They have an interesting partnership with Symantec and their Veritas Storage Foundation product. As you delete files the volume manager communicates with the storage array telling it what blocks have been deleted and can go back into the shared pool. Now you can get your 800GB of space back!
This is, in principal, similar to the Windows 7/Server 2008 R2 ‘trim’ command which tells your SSD drive which blocks have been deleted so it can do housekeeping and keep your SSD optimized. Maybe 3PAR can tie their thin engine into the TRIM command and bypass the Veritas volume manager?
For customers with a highly virtualized infrastructure, using ESX, I hope 3PAR and VMware can integrate their APIs so that thin reclimation works for VMs stored on VMFS volumes. ESX 4.0 supports changed block tracking, so I don’t see it as a huge stretch to integrate ESX with the 3PAR thin reclimation technology just like they did with the Veritas volume manager. *Cross fingers*