Is tape dead? No, but it isn't healthy. Enterprise Strategy Group (ESG) recently completed a comprehensive research project to see just how many IT shops were considering replacing (or already replacing) some or all of their tape infrastructure with disk. Note that I said replace, not augment. The results are startling.
Seventy-six percent of the respondents said they've already replaced, or will consider replacing, tape within 24 months. In fact, the mean capacity replaced by early adopters is 50TB to 99TB. Looking forward, the majority of those surveyed expect at least 40% of their current tape-based capacity will migrate to disk over the next 36 months.
The organizations surveyed for this research were "big dogs"--those with more than 10TB of production data on the floor, and most were much bigger than that. While the midmarket shops certainly buy their fair share of tape devices and media, paradigm shifts such as this tend to be driven from the top down, so we started at the pinnacle to gather our information.
What does it mean? Well, for one, you now know why tape companies have entered the disk business or are looking for new jobs. StorageTek has been in the disk game for a while, but it needs to shore up its library beachhead with solid offerings to maintain its revenues/footprint as tape library usage diminishes. Quantum was an early pioneer of virtual tape libraries (VTLs), and Overland has been shipping hundreds of disk-based iSCSI virtual tape units every quarter. Spectra Logic is adding disk inside its large libraries, while ADIC is figuring out what it all means.
My contention is that people will first back up to VTLs, disk systems that look like tape drives/libraries. They'll do this because it makes backup happen at disk speeds, which are much faster. This is sort of a write cache in the backup process. The data will then migrate from the cache to ultra-dense disk storage, the kind that is: a) massive and b) probably very inexpensive. It would have to be priced right because if I replace 50TB of tape with disk, I want it cheap.
This is a very interesting area for market and technical reasons. For the market, it's a new opportunity for the disk guys--it represents a possible 250%-plus niche (if a company has 10TB of production and 50TB of tape, they'll add 25TB of disk to replace the tape capacity), and that's just the mean of the Fortune 1000; the average numbers are much bigger. Technically, it offers those who can make 25TB fit into a box that has only 2.5TB--through compression, de-duplication, coalescence, etc. (technologies that write unique blocks only once, as opposed to a zillion times)--a boon economically as they can now offer 25TB of capacity for the cost of 2.5TB plus some nifty profit. Regardless, it could lower the cost of disk subsystems by an order of magnitude.
From there, data will be systematically written (in its entirety, and pieced back together as a whole) to tape, but nowhere near as frequently as we do today. We'll need only one full backup per week (if that), which we'll then send to a bunker for doomsday defense, and no incrementals (they'll stay on the ultra-dense stuff). This means you'll need 90% less media than you currently use. If you use 100,000 tapes per year, like a lot of storage people do, that represents a cool $4-plus million per year in media savings alone--which will buy much more disk. Yikes!
Tape will be relegated to its original intent, the deep offsite archive. Some organizations are talking about replacing all of it and using replication to move disk-based stuff offsite, but I think we'll be keeping full tapes around for a while. Dick Egan (the founder of EMC) always said no one needed tape. It's funny how that seems to be coming true just as EMC has done a deal with ADIC.