I have to design and determine the partiton size for Windows and SQL 2005 .Please suggest the needed paritions with mirroring option (or) general recommended one.
I would appreciate your help.
Alberto Castillo replied
You already have redundancy on disks, Usually with High Availability systems, disks are on a SAM (or similar) and the servers connect to it so when one server goes bad, the Cluster takes on, and the service (SAP or whatever you need) keeps going after a few minutes (Active- passive) or in no time (Active- Active).
As you can see, disks on arrays are not that easy to get bad and if so, you can replace the bad one and keep going, in the other hand, if the server (hardware) goes bad then you need another one to replace it but in your case you would need to take the disks out from one server to the other.
If any system have to write 4 times then performance will suffer a lot, for RAID 1 disks the system has to write all info twice that´s why it is not used for databases, now imagine if the system have to write it 4 times. In RAID 5, info is written with parity so if something goes wrong with one disk, the data can be reconstructed but calculating parity and segments to write also consumes time on the CPU so if you multiply by 2 then CPU will suffer too.
If you want, you could have a second server so you could do backups and save them to those disks so it is faster to backup or restore, in the event of the PROD server to go bad then you could use the same server (if they are the same) to move your disks and get everything up and running in a minimal downtime.
New servers (most models used for SAP) have memory redundancy, disk controller redundancy, power redundancy, etc so it is very difficult to have a major failure on those systems.