On 5/02/2016 5:15, Paul Gilmartin wrote:
I mentioned Java. Not realistic. It produces correct answers, but
performance is abysmal.
Do you say that because you tried it, or because everyone says it?
As part of the development of the API I have been doing some performance
testing. These are CPU times to read/report about 8GB of SMF data:
DFSORT:
CP time 0.78 s
Assembler, reading every record but no reporting:
CP time 1.24 s
C reading every record, no report:
CP time 1.4 s
Java batch job, basic report:
CP time 0.9 s
+ zIIP time 2.4s
(These shouldn't be considered accurate benchmarks, they were tests to
get a basic feel for performance.)
Java is not as good as assembler, but I wouldn't call it abysmal. I have
ambitions to improve it although that may be difficult as most of the
time seems to be in the I/O routines.
Regards
Andrew Rowley
Black Hill Software
+61 413 302 386
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
