As storage and computing professionals, we know all too well that moving disk drives between different RAID controllers is a royal pain in the neck -- if we want to preserve our data, that is.
One of the great ironies in storage is that even though disk drives and RAID controllers follow strict interoperability standards, there's no standard dictating a common data format. Move the disks away from that specific (or similar) RAID card, and your files become unreadable.
The reason why is just as frustrating: Vendors have no incentive to ship and no interest in shipping interoperable RAID solutions, so they build custom controller logic, which makes the controller's data content meaningful only in that context.
So how do we manage to not lose data during migrations? By doing a backup and a restore, or by any other copy mechanism that pushes data through the target controller. Unfortunately, those operations take a long time -- time often stolen from your company's working hours.
In a previous job I had to explain this unsavory technicality to an irate CEO who wanted to know why it took so long to move data from server A to server B. Didn't we buy all our servers from the same vendor for compatibility and peace of mind?
After explaining that although the servers came from the same vendor, our two RAID systems were incompatible, I had to calm down the man and clarify that dropping the vendor from our grace wouldn't do any good because, well, see above.
I'm remembering all this now because we should soon get some relief. The Storage Networking Industry Association (SNIA) has made RAID compatibility its target for quite some time, promoting the idea of a common RAID format as a basic interchange structure among different vendors.
The official SNIA acronym for this is DDF (Disk Data Format), and we should see the first interoperability tests this fall. An updated draft of the DDF specs was made public this week. Among other things, it adds support for RAID 6.
To quote from that document, "The Common RAID DDF structure benefits storage users by enabling data-in-place migration among systems from different vendors." This, I may add, sounds much better and faster than the alternatives I mentioned before.
Now if only we could persuade vendors to use compatible disk sleighs, to simplify (or make possible) moving a spare drive from, say, an HP machine to an EMC array. Don't hold your breath -- I'm not aware of any SNIA project to standardize that, but there is other spec news on the horizon that you will find interesting.
First, as you might expect the week before Storage Networking World Spring, is a reminder that full compliance with SMI-S (Storage Management Initiative Specs) is expected from all vendors by year's end. In addition to that first compliance level, which essentially targets basic services on FC (Fibre Channel) networks, the SNIA released new specs, dubbed SMI-S 1.1, that extend the scope to a wider universe of management services, including file servers, iSCSI, and tape libraries.
Interestingly, a third version of the SMI specs is aimed at application-aware storage management, but that project is still in its infancy. For now we should focus on the year-end deadline and let vendors know that we're watching and waiting.
If you have a suggestion for one of the upcoming projects, let your voice be heard at SNIA or your SNIA User Group.