In 2009, there were quite a few “I told you so” bragging moments by many mainframe advocates. Contrary to analyst predictions, there was no millennium meltdown of Big Iron computing. There was, instead, a healthy recommitment to mainframes by many business and IT decision-makers who had long planned to decommission their rigs by now. Owing in no small part to the superior story the platform had to tell in the strategic areas of cost-containment, compliance, continuity and carbon footprint reduction, the mainframe received a new lease on life that will, by all estimates, carry over into the next decade.
I hope business leaders and IT mavens will keep this reversal of fortunes in mind as they consider other strategies in the coming year, including but not limited to, the abandonment of tape technology and the adoption of nascent clouds. We can’t afford to let third-party “guidance,” whatever the reputation of the source, get in the way of good, old-fashioned common sense.
On the first point, tape technology (like mainframe technology) has been bad mouthed by just about every industry analyst who derived some part of his or her income from disk array vendors over the past few years. We’ve been told that tape is slower than disk (not true; streaming to tape is the fastest magnetic write modality ever invented), that tape failures occur with increasing frequency (not true, especially with advances in modern libraries), that tape isn’t keeping pace with disk storage in terms of capacity (interestingly, tape densities are much greater than disk in the same footprint of raised floor space), and that tape is too labor-intensive compared to disk (nonsense; tape is increasingly automated).
Now, pundits are raising the red flag of shrinking availability, what with Sun Microsystems being swallowed up by Oracle (a few wrinkles there) and with them, enterprise tape maker, STK. With IBM being the last Big Iron class tape vendor standing, the pundits are saying tape’s dance card is looking pretty thin. And, the cautionary fingers wag; Big Blue’s heart may not really be in it given that enterprise tape is a “replacement market” at best with little hope of growth.
This advice is being proffered to my clients with increasing frequency, despite the fact it’s full of horse-feathers, mostly by shills for disk vendors who seek to sell multi-hop disk mirroring as an alternative to streaming tape. In multi-hop mirroring, the array vendor sells you three copies of his overpriced disk arrays, a couple hundred dollars of synchronous and asynchronous replication software, and a WAN interconnect between array 2 and array 3. It’s important that all three arrays be of the same type, brand, and configuration or the software won’t work. Even if they give away the hardware, the software licensing fees create a cost of ownership curve resembling a hockey stick—not so with tape.
Analysts and vendors offering the disk alternative forget to tell you that disks die and mirrors are difficult to verify unless you stop their operation to check whether the data you think you’re replicating is actually being replicated. No one likes to break their mirrors to validate them; it’s a buzz hassle. But the truth is that data volumes have a tendency to move around and the disaster recovery mirroring process is often the last thing anyone bothers to update with new configurations.
Now I’m beginning to see more cloud woo enter this discussion. Proponents of clouds are suggesting their preferred computing metaphor drives a stake through the heart of already feeble tape. I’m not buying that one, either, except perhaps in the context of the small shop. Clouds, as presently conceived, are mostly just another standards-free lock-in and one in which the cost to ingest data is usually as inviting as the cost to retrieve data once stored is discouraging. Be sure to do the math.
I think tape is still viable and cost-effective. It’s a resilient and intelligent backup medium when used sensibly and with careful attention to sorting out the junk drawer of storage to identify data assets that need to be quickly restored in the event of an interruption event. From a CAPEX and OPEX perspective, a well-run tape environment rivals any disk environment out there.
Having just returned from SuperComputing ’09 in Portland, OR, where I saw Spectra Logic’s new tape library, T-Finity, I’m emboldened to say that tape has a lot of runway ahead of it. Don’t take my word (or anyone else’s) for it; run the numbers yourself. Like mainframes, tales of the demise of tape are greatly exaggerated. ME