Tuesday, August 28, 2012
The mainframe is alive and well
The mainframe is alive and well, according to a Business Day article by Steve Lohr in the New York Times on Aug. 28, 2012, “I.B.M. mainframe evolves to serve the digital world”, link here.
Companies are still buying or leasing mainframes, because of their reliability, security, and enormous throughput. The article discusses Primemerica's recent acquisition
IBM’s latest model is called the zEnterprise EC12.
In the late 90s, OS390 was the rage.
Do programmers in IBM shops still maintain JCL and complicated procs? It was the verbose JCL that was considered difficult to learn back in the 1970s, when Univac 1100, by comparison, offered a syntax a bit like today’s Unix.
Throughput improved enormously in the mid 1990s. Random VSAM updating was a problem with a “print stacking” application that I implemented 21 years ago yesterday in a salary deduction billing system, as it could take an hour single thread to update 30000 print image records. (We called the system affectionately, “Bill’s bills”.) Things rapidly got much better by about 1995. But mainframe batch programmers learned that you could process much faster by sorting and processing sequentially.
In a previous decade, being able to sort quickly was an accomplishment. But at Chilton Corporation in Dallas (now Experian), we could sort 200000 300-byte VL records in about two minutes, and that was considered an accomplishment. By the late 1980s, at a company called Healthnet in Richmond (Blue Cross) on an IBM 3090, we could do something like that in about thirty seconds wall clock. But the 4341 and 4381, which smaller companies (like Lewin then) bought, processing was much slower.