Dangerous Hardware 2489

The IBM MSS a.k.a. 3850 was a large tape robot (large storage for many tapes, with quite a few tape drives to read them, and a large robot that moved tapes between the storage bins and the drives). In the MSS, the tape cartridges where cylinders a little smaller than a beercan, sitting in honeycomb-shaped slots in the two walls of the MSS. The MSS itself was about 2m=7' high, 1.5m=5' wide, and 6m=20' long; the two walls were covered with tape cartridges, with a narrow aisle in the middle, where the robot moved. For the life of me, I can't remember the capacity, it must have been HUGE - hundreds of gigabytes :-) :-) :-)

The MSS was used in conjunction with several disk drives (3350 or 3330, in my vague recollection) to provide virtual disk service. In a nutshell, the MSS pretended to be a disk with thousands of replaceable volumes, which were automatically replaced when accessed. At my installation, the MSS was tied into an HSM system (hierarchical storage management) called FAST: it kept the most-used datasets (the thing we would call a file today) on disk (3350, later 3380); more rarely used datasets on the MSS on virtual 3330 or 3350 volumes), rarely used datasets on 6250 bpi reel tapes (mounted by human operators) that were stored in the computer room, and the most rarely used datasets on 6250 reels in a warehouse. Access time to the datasets ranged for milliseconds for disk to seconds for the MSS to minutes for machine-room tapes to days for the warehouse, but all automated without user intervention.

The robot inside the MSS was large and powerful, and moved on a track. Part of the field service procedure was to disable the robot before entering the narrow corridor. There was at least one bane accident, where a field engineer entered and was crushed by the robot. I don't know how many injuries and near-misses existed, but we were all extremely respectful of the MSS.

At my site, the only mishap we had was the following. One night, one of the tape drives in the MSS failed, and any tape that was mounted on it was marked as "unreadable" by the system. Our HSM system automatically placed unreadable tapes into the output hopper, so that a human operator could deal with the problem. Unfortunately, every time a tape was wrongly declared unreadable, the batch job trying to access that tape would crash, and another was started. After a while, all the tapes had been tried and rejected, and thrown out. As the output hopper on the MSS only had limited capacity, it overflowed. In the morning, the day shift operators came in, found the MSS emptied of tapes, many thousand cartridges rolling all over the floor, and most of the batch queues drained. The users found that for a day or so, the MSS had to be shut down (fundamentally making the whole computer center useless) while all tapes were manually placed back into it, and then they had to resubmit their batch jobs.

-- The address in the header is invalid for obvious reasons. Please reconstruct the address from the information below (look for ).