I've been having a correspondence with Dave Truslow of Rocksoft. He thinks that MAID arrays may be being used literally like sequential tapes because their structure inhibits random I/O. (His argument is included below.)

Dave writes:

"MAID works by keeping a subset of the entire array online. This reduces heat and power and allows denser packaging. I believe that Copan's design limits the number of online disk to 25 percent of the total array disks.

"In most cases, the array is divided into groups (MAID group) of disks that can be concurrently powered. These typically align on RAID boundaries too i.e., a MAID group is one or more RAID groups.

"This means that one must be careful in file accesses crossing MAID cluster (the number of concurrently powered disks) boundaries. For example, Group 1 and Group 3 may be concurrently powered or Group 2 and Group 4 may be concurrently powered. I will be unable to access a file system that grows beyond the limits of Groups 1 and 3 or 2 and 4, nor can I create one that uses Group 1 and 2 or 3 and 4 since they cannot be online at the same time.

"A MAID set can be used for random I/O such as a database. However, this quickly becomes difficult to manage as file system sizes grow. At most, the MAID array's online capacity is limited to the maximum number of concurrently powered drives (25 percent for Copan I believe). These considerations become messy and in my experience most customers simply use them as sequential devices as tape replacements. This is much easier to manage."

And is MAID nixed anyway?

Dave also asks this question: "Does MAID have a compelling value with the advent of disk in a cartridge?"

This is in response to Imation's product which uses the idea of having disk drives embedded in tape cartridge form factor cases. The idea is that they are stored in a library and robotically fetched to and from disk drives. The whole device looks and behaves like a tape library but part of it actually uses disk technology.

Of course, it doesn't have the storage density of a Copan MAID or Nexsan AutoMAID-based product. It would also have a longer file access time, I think, as disk-in-a-tape-cartridge fetch time would be longer than idle disk spin-up time in a MAID array.

But it looks like a good way of adding disk-based archiving to an existing tape library in an easy fashion. You can take the disks out for off-site vaulting, which you can't do with a MAID array.

It's a conundrum and, as ever, the market will decide. I'm fascinated to see the first tape library products emerge that use the Imation disk-in-a-tape-cartridge technology.