Re: onstat -g ses

Dagmar Wartenberg wrote: > > Vic Goldberg schrieb in Nachricht <3496BD2C.74BB@cornell.edu>... > >dalink@west.com wrote: > >> > > > >Also, look at PSORT_NPROCS environment variable. Controls number of > >sort threads, and can really speed things up. > > Two weeks ago I participated in a training about performance tuning > for ODS at Informix Munich. The referent told us that PSORT_NPROCS > can still by found in all ODS manuals but it had no effect anymore. > Even without setting the variable in the environment the sort proces > would work as fast as possible. Anyone out there who will coment this? > Is this true? I'll challenge that. As late as 7.21UC3 PSORT_NPROCS still improves index builds and presumably other large sort jobs. I have not tested it in 7.24 (which we are evaluating) and 7.3 (which we are Beta testing) but I suspect nothing has changed. The docs do indicate that the PSORT* variables are no longer effective and that is the common wisdom among Informix trainers and consultants but experience tells. Try various combinations of the follow yourself and see: PSORT_NPROCS=40 PSORT_MAXALLOC=10240 PSORT_DBTEMP=<at least 3 filesystems with approximately the same free space available as the smallest will limit usable sort space> PDQPRIORITY=20 Create an index on a million row plus table with and without these four variable values and with any of them eliminated (like keep PDQPRIORITY but drop PSORT_NPROCS) or adjusted and time the job. You'll be surprised how well this "outdated" feature still works. I will do this formally myself in all three versions and post the results. Art S. Kagel