QTP Performance - Clarification
Ron Burnett
ron@cryptic.rch.unimelb.edu.au
Tue, 10 Jul 2001 11:24:28 +1000
Many thanks to everyone for their useful comments.
First, I should have made it clear that this is a batch-job. I take a dim
view of anyone running QTP interactively on our systems, except in
trivial record-volume situations or for testing bits of code.
The CM KSAM file sizes are not large; the one giving me trouble is
a code describing country of birth, and I need to convert to a new value
three times for each input record (one for patient, one for mother, one
for father). The first time I did this, it took 683 elapsed minutes, but
I was linking from the input file to the KSAM file three times (once
directly, twice via alias). There are a total of 289 records in the KSAM
file.
I've split the run into three jobs, each linking only once to the KSAM file.
The input file is 2422 bytes per record. I can't use QUIZ either to create
this file, or to output a separate updated version because the record size
is well beyond QUIZ's limit.
I'll try compiling the code, but my understanding is that the only penalty
for not doing so was at load time. Since there are only 4 statements,
how much benefit could one expect? Am I to understand that uncompiled
QTP runs entirely interpretatively?
I think the suggestion to convert the CM KSAM file to NM KSAM is also
worthwhile. Easily done.
Pre-sorting the input file is another worthwhile suggestion. Thanks Ken!
Yes, Matt, the file in the access statement is the same as the one in the
output statement, and my pseudo code should have read
item <new-code> final <t-new-code>
I can't tell you the timing of the input and output phases, but this leads
back
to my original suspicion .... I haven't got enough memory to process such a
large file efficiently. The idea of generating another output file is
probably
going to help more than anything else, even at the cost of disc space (always
an issue, but manageable).
I'll do some experiments and let everyone know the results.
Thanks again for all your help!
Ron Burnett
ron@cryptic.rch.unimelb.edu.au