Our env is OLTP + DWH 24/7. Applications goes from financials to manufacturing and OLAPs. We run EVERYTHING you can name it on 1 giant SAN.
The current engineering thinking, that is strictly from an engineering sense, is the more disks pushing the load the better.
That is true but it can be even better if we segregate some basic areas is what I believe.
You will still have that many disks active but they will work more intelligently. The current thinking is like - trust me this big pool of disks will work . Yet we get constant glitches of peaks,locks, i/o waits exceeding max, and waits on transaction logs...etc. Then it goes away for a while then starts up again. Application codes are one area but I am thinking more systemically on segregating the SAN using dedicated spindles (forming a RAID group or Storage pool) for:
data
index
tempdb
tran logs
If i have to compromised maybe combine data + index for 1 physical RAID, Tempdb for another and tran logs on yet another.
Am I old school, that new DASD with automated load balancing and cached memory and solid state channels make it unnecessary ?
Is your set up like what I described or "trust me one huge SAN is way to go" ?
Thank you in advance ``RL
RSL