Guidelines for IMS system data set placement

Here are some guidelines you should consider when placing your IMS online data sets:

• Separate among lightly used DASD volumes: heavily used libraries and data sets such as IMS.PGMLIB, the online log data sets, the active IMS.FORMATA/B and IMS.ACBLIBA/B libraries, and message queues.
• If high-activity data sets are on the same device, place them close to each other. The space allocation should stipulate contiguous storage and allocation by cylinder to avoid fragmentation of stored data.
• Have the volumes available to IMS mounted as private so that operating system temporary data sets are not allocated to these volumes.
• Do not place online IMS system data sets on shared DASD volumes. Avoid contention for both the device and the control unit. Position primary and secondary OLDSs on separate control units and channels when possible.
• To optimize program fetch I/O, order PGMLIB in descending frequency of use. Use full-track blocking. This minimizes Start I/O (SIO) operations and seek times.
• If using fixed-head devices, place write-ahead data sets (WADS) and database index data sets on the fixed-head portion.
• If program library contention is a problem, create multiple copies, each for a subset of the message regions. Do not mix production and test program libraries.
• For IMS shared data sets subject to updates, place the data sets on DASD devices capable of allowing parallel I/O from the same or multiple systems.