When trying to create a new PDS, I have the primary and secondary allocations both set to 10 cylinders.
The data set itself is an FB 255 but there's one parameter it's asking for which has me flummoxed.
That's the block size. This has defaulted to 27920 but I'm wary of using that value since it also defaulted to FB 80 and I'm worried these two may be related somehow.
How do you calculate a good block size for the PDS or does it not really matter?
These days the best to do is to just not specify any blocksize anymore, e.g. within JCLs (batchjobs running on the mainframe). That way you leave it to DFSMS (and the people responsible for configuring DFSMS) to always use the best possible blocksize automatically. Same is true is you allocate a dataset from within (eg) a CLIST or REXX procedure.
That's also why over the years, many mainframe sites have performed all sorts of conversions, to just mass remove all such blocksize specifications (eg in JCLs).
It's a while since I've done mainframe work but basically blocksize has an impact on disk utilisation and disk response time. For FB it should be a multiple of the lrecl. Utilisation was related to how many blocks could fit into a track and how well utilised that made the track. For disk response time the recommendation usually came to making the blocksize approximately one half a track. This was probably from 8 years ago and in primarily a non-SAN environment.
