It's just past Y2k, and this big company is in the process of moving its data center to share space with an affiliate company, reports a mainframe pilot fish on the scene.
"The majority of the batch work at night was processing sequential files," fish says. "We found that the best solution in terms of cost was an automated tape library. The data transfer from tape was the same data rate as from disk, so the only overhead was an automated tape mount that took 20 seconds.
"The tape library was large enough to hold all our data for 35 days, so we could process our daily, weekly and monthly jobs with almost no actual human intervention, and it was considerably less expensive than disk space."
But in the run-up to the move, the company's CIO retires, and the affiliate company's CIO takes over. His plan: Keep the two companies' data separate to avoid legal complications, and install a new mainframe at the new site, with the existing mainframe decommissioned.
Fish notices that there's no mention of moving the tape library, so he asks how that's going to be done -- with new hardware, as with the mainframe, or an actual move?
None of the above, fish is told. Instead, the automated tape library will be replaced by half the space in the affiliate's virtual tape library -- essentially a disk array masquerading as tape drives.
"I informed them that a virtual tape system would not satisfy the needs of our processing," says fish. "I was told it would work.
"And we weren't given a chance to test out our actual processing at the new site prior to going 'production.' We were told it would work, after all."
So once the new hardware is in place and checked out, fish's team moves its data to the remote site and begins processing -- only to discover that, instead of 35 days' worth of data being available within 20 seconds, the virtual tape system's cache is only large enough to hold 22 hours of data at a time.
And when the right data isn't in the cache, the system writes data to a physical tape. Then another physical tape has to be mounted to get the requested data into cache -- and only then the job can continue.
Result: Nightly processing that used to take less than four hours now takes three times as long.
"Lots of data had to be moved to newly purchased, much more expensive disk space just to get the work done," grumbles fish.
"And why were we not moving the tape library? I asked. The CIO said he didn't like the vendor!"
You know what Sharky likes: True tales of IT life. Send yours to me at firstname.lastname@example.org. You'll snag a snazzy Shark shirt every time I use one. Add your comments below, and read some great old tales in the Sharkives.
Get your daily dose of out-takes from the IT Theater of the Absurd delivered directly to your Inbox. Subscribe now to the Daily Shark Newsletter.