Optimising MIPS: five changes with immediate effect

In the first days of a mainframe assessment you are not looking for elegant optimisations. You are looking for the ones you can make straight away, without touching the business logic and without renegotiating the batch windows.

The first checks concentrate on a handful of recurring cases. One: redundant sorts, usually inherited from chains processing flows nobody uses any more. Two: repeated reads of the same table inside one step, which a single read into an in-memory structure can replace. Three: processes run every night over data that changes once a month.

Four: scheduling. Moving some chains out of the peak can do more for the bill than a good deal of rewriting, because the cost follows peak consumption and not simply the total volume processed. Five: programs that read a lot of data only to discard most of it, where applying the filters earlier improves efficiency.

None of this replaces the long-term decisions. It recovers capacity and buys room for the structural work, from rationalising the chains to modernising the applications.