Re: performance problem
Posted in 1997
Sorry if this is a little basic. Have you unloaded / re-loaded any data. If yes, have you run an 'update statistics' lately ? Has anything else changed. Are you sure it's to do with configuration rather than the application ? Have you tracked down the individual statements causing the problem ? Mike. In article <5g49n7$rla@cssun.mathcs.emory.edu>, Michael Silver <MICHAELS@dartinc.com> writes >I think you mean that there are two instances of Informix running. We >have a K210 with four processors, with only one instance, and if we run >two batch jobs (lacking indexes) at the same time, the four processors >get pegged at 98-99%, and the rest of the users are locked out. We had >the same problem with two proc., and on an E. I am surprised that you >only now see a decrease in speed. One thing to try: set OPTCOMPIND to >0 (zero). This is the optimizer, and it will get out of whack depending >on the configuration and database structure. It may help. Also be >weary of PDQ, it has only caused problems for us. > >...Michael... > >}-----Original Message----- >}From: Jane Johnson [SMTP:johnson@holidayco.com] >}Sent: Monday, March 10, 1997 1:45 PM >}To: informix-list@rmy.emory.edu >}Subject: performance problem >} >}Hello all: >} >}We have a HP-UX B.10.01 E 9000/829/K400 8-user system. It's a >}multiprocessor machine (I believe it has two processors). There was >}2 informix server resides on the system. We recently add another >}Informix >}server on the system. We notice that one of the server's performance has >}gone down dramatically (jobs used to run in minutes now runs for hours). >}Besides adding a new server on the system, we also increased the >}SHMVIRSIZE to decrease the number of allocated new shared memory >}segments for one of the servers on the system (not the newly added the >}server). We also add a new disk (2g) to the system. Well, does anybody >}know what we should look for this performance problem. Thanks very much >}in advance. You can e-mail me at johnson@holidayco.com. >} >}Thanks again! -- Mik Hobbs