MVSFORUMS.com

[hellohatim]

Hi Sharon,

Here are the statistics again with Elapsed time as well. For first two Cost and elapsed time contain other steps as well so cannot include them here.

[jsharon1248]

For these files, you can fit the entire index into the buffers. Right now, you're way over allocated. You specified BUFNI=128 which means you allocated buffer space to store 128 index CI's. But looking at the REC-TOTAL values for the index components, the highest value is 6. I would match the BUFNI to the REC-TOTAL for this process. For other random access processes, you would want at least 1 index buffer for each index level. For these files, the LEVELS values are all 2 or less. The BUFND values are a little trickier. For true random access, you really only need 1 or 2. But I'm betting this process is closer to skip sequential, so I'd stick with the 16 you have now and run another test with 20 to see if that has a positive impact. You quickly reach a point of diminishing returns, so you want to review the results of different BUFNx values.

I remember asking you about storing the last read record for each file in working storage and then checking to see if the record you want to read is already in WS before issuing a read to the VSAM file. If you haven't done that, I'd recommend doing so. You might be surprised at how much you can save by that alone.

[hellohatim]

Hi Sharon,

Thanks for the information. Yes 128 is way too high, 6 should be more then enuff I guess. I will try these values once to check.