We are hosting a 140 GB database on SQL Server Version 7 and Windows
2000 Advanced Server on an 8-cpu box connected to a 15K rpm RAID 5
SAN, with 4 GB of RAM (only 2 GB of which seem to be visible to the
OS) and a 4 GB swap file. (The PeopleSoft CIS application will not
permit us to upgrade to SQL 2K.) We recently upgraded the server from
4 to 8 cpus and the SAN disks from 10K to 15K drives. But we still
have heavy SAN disk usage, sometimes at 100%, and read queues often
averaging 4 and peaking at 12.
The CPUs are loaded at only 20-50%. (The politics are such that it is
easier to throw hardware at the problems.)
We are looking into archiving, converting from RAID 5 to RAID 10, and
at splitting the mdf file into several file groups in an attempt to
get more disk heads into play. (We are also looking at rewriting the
application to reduce the read volume and frequency.) Does anyone have
any other ideas?
Incidentally, does swapfile get used when the physical memory equals
the OS maximum? If the OS can only see 2 GB and we have 2 GB (actually
4 GB) of memory, is the 4GB local swap file on the C drive unused?
Thanks in advance for any assistance.