The following is a guest post from Bob Schmidt, a performance engineer for mainframe DB2 at State Farm Insurance. It is a response to a SearchDataCenter story on how tuning mainframe applications can cut software costs. Anyone interested in writing guest posts on Mainframe Propeller Head can email firstname.lastname@example.org.
The article on mainframe tuning was OK. But it missed the real problem of mainframe cost. DB2 is cost effective when used as a relational database. It can also work as a hierarchical database, or as a flat file. But it is not cost effective in those other modes.
The problem is improper application design due to designers and architects not thinking in relational terms, nor in sets, but rather in sequential, procedural processing. Modern hardware and systems software, including DBMS, is intended for parallel processing which is best used with set processing and not with sequential processing. Yet too often, the architect thinks that parallel processing is sequential processing n-wide.
The real solution is good design.