Re: Service Program Overhead

Reading via a keyed access method is "random" I\O which can be significantly slower than reading via an arrival [i.e. sequential] access method; blocking can be specified for the latter.

If a VIEW can be used for the unordered access of data, then the SQL SELECT statement [directly or in a] VIEW could possibly even encapsulate the formatting logic against the data within the I/O buffer. As a SELECT with an ORDER BY the method for ordering can be left to the optimizer; e.g. a sort of sequentially read data or data read in parallel may be more optimal than keyed\random I/O.

Regards, Chuck

James Perkins wrote:

Well unfortunately I can't use blocking because I'm using SETTL,
READE and CHAIN.

I have not tried a view, but that may work thinking about it. If I get the time I might try it. I've already gone over budget due
to trying all this stuff out, but what are you going do ;-)

Nathan Andelin wrote:

I may be able to suggest a few performance tips. To begin,
are you using block(*yes) on the F-spec of your input file?
If not, does that improve performance?

I'd also be interested in hearing the performance difference
between reading an SQL "view", rather than your read + chain
sequence.