All, I have a datawarehousing environment with multipe db's up to 400gb and a table with 1.7 billion rows. This table is often involved in complex joins etc..... Env= SQL 2005 STD , windows server 2003, data over 3 data drives (non SAN) of RAID 10 storage with 10, 4 and 6 spindles respectively. Anyway , In perfmon I have I/O numbers that are off the charts: % Disk Time = 900 Avg Disk Queue Len = 50 -150 Page splits = up to 200+ Page life expectancy = 70 Any ideas would be appreciated. It seems this is a pure hardware disk issue as sql code is well optimized and tempdb is on dedicated storage. thanks much
You need to consider the factors such as IO throughput that are the number of spindles, and drive throughput such as sequential and random IOs per second. Have you taken up further counters on disk queue, as this will effect the queuing that will be reflected in high disk seconds per read or write. BTW is this a SAN? Data Warehouse loads can be either logged with inserts, updates or deletes, or non-logged using bulk copy.