Logical depop, doesn’t that sounds like a new type of pop music? Not quite. Technical committees T10 and T13 which are responsible for the interface standards for SAS and ATA respectively, are brewing up a new standard called Logical Depopulating or depop. There are two types of depop which are being drawn up, offline and online. The main idea of the offline depop function is to give a system the capacity to decommission a specific physical element of a storage device such as a disk head, SSD die or channel. Afterwards, the system would have the remaining portion of the storage device reformatted and put back into service with a reduced capacity. This does cause data loss on the drive and downtime associated with the process, yet also extends the drive life-cycle and reduces service times by delaying replacement.
One main point the T10 committee addresses beyond complete drive failure are performance issues in a redundant storage environment. When one drive takes an unusually longer period of time to respond than the other drives, the request will end being only as fast as the slowest drive. This is particularly concerning when the host software may already have access to the information the user is requesting due to redundancy, yet due to the way storage interfaces currently work, the host software must needlessly wait.
According to Joe Breher, a storage architect who is leading the standards effort, “We’re seeing in a large percentage of cases in drive failures, when they do finally fail, the failure is limited to a single head in the device. What can we do to take advantage of that fact and extend the service life?”. You couple this with the fact that most enterprise drives are now also sealed devices so particle contamination of the entire device is less common.
The online version of the depop will eliminate the requirement to reformat and the only data loss will be in the decommissioned elements. Essentially, some new commands will be introduced for the drive and host to determine which surfaces(Logical Block Addresses) to depop, so the affected ranges can be deallocated and then have the device run in a thin provision mode. The affected elements may be truncated, amputated or regenerated depending on the circumstances in order to restore operation as a fully provisioned device, but with a lower storage capacity.
Due to the way filesystems have been used over the last several decades, it will take quite some time for the changes to mature so that this can be a usable storage interface standard. Microsoft intends to deploy an offline version within its Azure platform in approximately 18-30 months and Western Digital’s Hitachi Global Storage Technologies where Joe Breher works and Seagate Technologies are both shipping devices with a proprietary version of the offline depop implemented. The question that may need to be asked is, how long will these ‘injured’ drives last even if they can be salvaged? What kind of unknown side effects will the filesystem changes produce?