Re: 4 disks. Best utilization?

In article <3295894f.10249224_at_n5.gbso.net>,

   chuckh_at_dvol.com (Chuck Hamilton) wrote:
>>also, what is the importance of down time? (you don't really have a
>>enough disks for a failure-proof system).
>
>Who does?
>
>Given a big enough catastrophe, 50 disks may not be enough. ;)
>
>For minimal fault tolerance you'd need:
>
>1 for tables
>1 for indexes
>1 for temp tablespace
>1 for rollback segments
>2 for control files
>2 for online redo logs
>2* for archived redo logs

Huh?? In an "ideal" system maybe! The control files need to be mirrored, but are not heavily hit. They can easily be put on the same disks as the redo logs without hurting performance.

Also, how do you expect "fault tolerance" with only one disk for data? I guess it depends on what you define "fault-tolerance" to be!

The last "kinda-fault-tolerant" system I installed (on Solaris) used Raid 0+1 for everything except the archived redo logs - which were simply mirrored. All this was done using a disk array and the Veritos software with 12 2GB disks for the primaries mirrored to another 12 2GB disks as the secondaries for the stuff on Raid 0+1. Another 2GB disk was used for the archived redo logs and another yet for its mirror. (Be careful about how all the disks are mirrored so that pulling a drawer or losing a controller does not bring the database down!) Then there were still a few disks left over to use as hot spares. I do not consider this system to be absolutely "fault-tolerant", but is fairly safe from disk failure. To make it more fault tolerant, you would need to add another Ultra-Sparc onto the other "tail" on the array and some additional software for process fail-over from one server to another. Received on Wed Nov 20 1996 - 00:00:00 CET