Re: OnLine Internals and Performance Questions

jimlenahan@aol.com wrote: > My environment is IFMX 7.20 on a 4 way HP K series system connected > to two HP G class machines. On the G's are up to 600 users running > applications thatneed to talk to the data base via standard IFMX net > connections. At peak times, user performance drops dramatically (what > should take scores of seconds takes scores of minutes). The only > noticable constraint appears to be waits for polling threads. In trying > to fix the problem, the following questions arose: > > - The IFMX docs say that MAXCPUVP should not exceed the real number > of processors in the system. However, my CPU's are typically running at > 50-65% utilization. Is there really anything that could go wrong if the > number of CPU VP's were greater than physical processors? > (My 7.1 experience is on Sun, but I don't think this would be different on HP) No, trying MAXCPUVP > number Processors won't hurt. But it shouldn't help, either. If you have more than 1 CPU VP thread on a processor, you'll just be increasing contention at a thread level. If you're not maxing your cpus, then the existing cpu vps aren't being kept busy - you've got to have another bottleneck. You may not have enough poll threads to handle the load, or something else. Are you waiting on I/O? Vic -- Victor Goldberg Cornell Theory Center 520 Rhodes Hall Ithaca, NY 14853