Re: 7.x performance issue on DEC Alpha

Hi Rakesh. I do not understand what you meant by the Alpha2 uses the Alpah 1's database. Are these databases on two machines cross mounted under NFS ? Anyway, since you have one of those 2 CPU machines I am including a mail that was sent to this news group specifically about the 2CPU machine problems. Run this by your Informix consultant and give it a try maybe this solves your problem. Regards, <<<Saeed>>> May 1, 1996 Subject: Re: System tuning > >This is on a system that has between 20-35 informix users, 128Mb RAM, >2x88100 CPU's running DG/UX5.4R3.10 The database is on 5x1Gb RAID5 >disks. Does it seem immediately obvious to anyone as to what we might be >missing here? Our system DOES NOT run as fast as it did with 5.x Our >users aren't terribly happy these days :( We recently encountered similar sounding performance problems using Informix 7.11UC2X?? on a HP K400 with 2 processors and 1280M (!) of RAM. We had almost 99% cache hits, with still extremely sluggish performance. Our final performance tuning victory came after playing with the NUMCPUVPS, MULTIPROCESSOR, and SINGLE_CPU_VP parameters in the onconfig file. MULTIPROCESSOR 0 NUMCPUVPS 1 SINGLE_CPU_VP 1 gave us stunning performance gains. The manual (in the very fine print) affirms that this is the way to configure our system with two CPUs even though the MULTIPROCESSOR 0 sounds a little funny at first. Informix attributes this to the fact that with a single CPUVP, the whole mutexing/latching/locking scheme is dramatically reduced in complexity. Best of luck... ------------------------------------- cut your mail ------------------ > >tolly brumbelow <tolly.brumbelow@bridge.bst.bls.com> writes: > >> Are you running the same application on both Alpha 1 and Alpha 2 ? >>Have you checked to see if swapping is taking place on the unix side? >>(if so you may have the buffers set to high) >>Are you running update stats on Alpha 2 on a regular basis? >>What does your onconfig file look like? >>Check to make sure the temp dbspaces have the temp flag turned on? >>(onstat -d and look for the T flag on the tempdbspaces) >>Are the users accessing the database the same? >>Run the queries with set explain on to see what path they are taking. >>Make sure your update stats follows the guidelines in the performance > > >We are running the same application on both Alpha 1 and Alpha 2. >There is very little swapping going on unix side. >Update stats is run regularly on Alpha 1 and since users on Alpha 2 >access the database residing on Alpha 1 no update stats are done on Alpha 2. > >Temp dbspace have the temp flag turned on.. >Users on both systems are accessing the database on Alpha 1 >WE had installed a probe to collect the queries during the day >and have run them so see what path they are taking and have already >made adjustments to all the important queries and we are doing update >stats based on the manual and as per Informix's consultant's advice. > >Some of the important onconfig parameters are : >Locks 12000 >Buffers 8000 >Initial Shared Memory segment 300,000 >additional shared mem 80,000 >MULTIPROCESSOR 1 >SINGLE_CPU_VP 0>Number of CPU VPs 1 to start with and then one is added dynamically >Kernel AIO >Informix AIO VP is currently set to 4 but we plan to reduce it to 1 or 2 >over this weekend >OPTCOMPIND 2> >onstat -g sch shows that the second CPU vp has 50% of it's cycle in >Busy wait....> >and if I reduce the CPU vps to one same queries start to take long time >although sqexplain shows the same execution path..To me it looks like >as if informix is generating major contention for all it's queries. >A query which would take 2-3 seconds with 2 CPU vps will start to take >40-50 seconds with 1 CPU VP even through it is follownig index path... > >If you would like I can post the actual onconfig file and the outputs >of onstat -g sch , iop, glo etc... > >- Rakesh