Using the DIRECT keyword

If you specify DIRECT(N) and the Recovery utility creates a log extract data set or CA extract data set, the recovery tasks read log and CA data from these data sets during the second recovery phase (in which the data is written to the output database data sets). DIRECT(N) helps to improve task overlap because the utility does not need to pass the data buffers directly to the individual recovery tasks.

If you want to take a more gradual approach to using extract data sets, you can use the default DIRECT(Y) keyword. With this keyword, the utility can create extract data sets in support of the Automatic Restart feature, but it can still pass the data buffers directly to the recovery tasks. If the utility is not creating extract data sets, the utility forces use of the DIRECT(Y) keyword.

For a failed job step that uses the Automatic Restart feature, that uses extract data sets, and that is restarted, the utility forces the use of DIRECT(N).