Fast Path spill files

Fast Path spill files buffer the log input records when parallel log readers are assigned and Fast Path input is processed.

You can avoid Fast Path spill files by setting the number of parallel log readers to 1 or by using the FP.NOSPILL command.

The maximum number of spill files is the total number of log files minus 1. Fast Path spill files are processed and released in sequence as the log files are read. Thus, the number of spill files that exist at any time is different from the total number of spill files.

Because each log starts and stops independently, you cannot be certain when each Fast Path spill file will be released. The working set of spill files can range from 0 to number of logs minus 1, but is generally the number of parallel log readers when parallel logs is greater than 1.

The size of a spill file depends on the size of the largest input log, the DEDBs to be processed, and the distribution of log records. Each spill file is created from a single input log.

The spill file must be large enough to hold all Fast Path log records for all DEDBs to be processed. A Fast Path spill file is, generally, not larger than the largest log and is frequently much smaller.

However, because IMS HP Change Accumulation Utility keeps additional information about each log record, if every record of a log file is a Fast Path record for a DEDB to be processed, then the spill file might be slightly larger than the input log file.