Re: [MI400] Read and write performance Records VS Fields

[Gary Guthrie <garyguthrie@xxxxxxxxxxx>]

The reasonable response would have indicated that the question was too
broad and general for a specific answer. In other words - it depends.

Regarding the O/S handling caching correctly - caching may not even be
an option to the O/S, depending on access method and such.

The reasonable response would also mention hardware rather than software
issues. Disk access involves actual moving parts. In other words -
relatively slow.

I think it reasonable to say that it is basically preposterous to think
that 100 reads of small amounts of data would ever perform BETTER than 1
read of all the data. At most, I'd say performance characteristics would
be identical. Unless, of course, the code that governs the movement of
data (and I suppose device microcode for DASD, too) is written such that
it's goal is to devise a scheme where 100 reads would perform better
than 1 for whatever perverted reason!

Gary Guthrie

Leif Svalgaard wrote:
> 
> well, it is the same number of bytes to be read, so if the
> O/S handles caching correctly it shouldn't matter (as the
> time to make 100 calls vs. one call can be neglected).
> But the original question was a bit silly (sorry sas).
> Without information about record sizes, field sizes, how
> many times the program will be run, etc., no reasonable
> answer can be given.