I tend to agree with Anjo on this.
You cannot beat the physics. SAN has the same kind of drives you would use on a system locally. Most outer part of the drive works about 2 times faster (for 3.5 in drives) than the most inner part. With a complete mixture of the storage area in a SAN your bottom line performance at best as fast as your slowest component can performs (the most inner part). Basically, you are slowing yourself down for no reason. Yes, you can get temporary performance spikes with SAN on a system that was not properly designed by reading some data from SAN cache. However, you are better off spending money on system memory rather than SAN cache (the first one is much cheaper this days!!) and use it to keep frequently accessed objects even closer to the database engine than a SAN can do.
128Gb of cache sounds a lot... not really for 20Tb of storage ;-)
Another example of SAN overhead is connection to the server. 2Gbit Fiber channel sounds fast but realistically give you 200Mbyte/s throughput. Only two 15K RPM drives at 100Mbyte/s (on average) transfer rate are enough to suck life out of a single FC. How much does it cost to install enough FC cards for your SAN storage to match all the drive performance?
Check this link out and do the math: http://seagate.com/docs/pdf/datasheet/disc/ds_cheetah_15k_5.pdf
Good luck, Sergey Popov
Anjo Kolk wrote: > Normally when things run slow, you are using a SAN. If things are > fast, you are local disks. If that really is your expectation, then I think you need to look at your SAN configuration for performance issues. Generally the access to disks on a storage array are faster than access to local disks. Write-cache is wonderful stuff when you can get it. > On 9/29/06, *J. Dex* <cemail_219@xxxxxxxxxxx > <mailto:cemail_219@xxxxxxxxxxx>> wrote: > > By using df -k, how can I tell whether a system is using a SAN > device versus > having everything on local drives? If it is all logical drives, > wouldn't > all of the mount points start out with something like > /dev/lvm0/ and then be sliced up among Logical Volumes? > J:
Which OS are you using? I am assuming Linux or HP-UX given the mention of LVM. The specifics would be different for each OS, but you should be able to trace from the device mounted (seen using df -k) to see actual controllers in use and this would tell you whether you are using a SAN or local disks.
Nathan Dietsch -- http://www.freelists.org/webpage/oracle-l