Re: index build performance
Posted in 2007
Floyd Wellershaus said: > Back on that 1.1 billion row table again. > Thanks to help from sources here, I was able to get the named pipes > working to half the time for a load and unload for the table. > Now I'm concerned with building the indexes. > >>From sources here also, I am going to bounce the server and use the >> following onconfig guidelines. > BUFFERS 25% of available memory > SHMVIRTSIZE 75% of available memory > CKPTINTVL 3000 (50 min) > LRU_MAX_DIRTY 80 > LRU_MIN_DIRTY 70 > RA_PAGES 32 (16 for 4k page) > RA_THRESHOLD 30 (15 for 4k page) > DBSPACETEMP Lots > DS_TOTAL_MEMORY 90% of SHMVIRTSIZE > DS_MAX_SCANS Nbr of fragments of largest table > PHYSFILE Large > > Are there any other tricks to improve the performance ? > Is there an easy way to estimate how big to make the PHYSFILE and the TEMP > DBSPACES ? Are the tables fragmented? If so, use PDQPRIORITY=50 and PSORT_NPROCS=4 then set PSORT_DBTEMP=/tmp1:/tmp2:/tmp3:/tmp4 where these are preferably on different spindles. If you have fewer than 4CPUs, set PSORT_NPROCS= the number of CPUs you have and reduce the number of directories in PSORT_DBTEMP. -- Bye now, Obnoxio "I'm astonished anyone pays real money for this crap." -- Cosmo -- This message has been scanned for viruses and dangerous content by OpenProtect(http://www.openprotect.com), and is believed to be clean.