If your primary goals are performance and redundancy therefore best cost unaware raid must be raid 0+1
This gives you max write performance on a zero stripe with full mirror for redundancy.
If your really serious about true production get rid of the SATA as MTB (Mean Time Between Failure)is not as good as SAS, better perf and life over SATA, 3GB throughput on the decent drives.
Dont put IDE on a production server if you can help it as failure rate is high, this machine is a 24x 7 machine I take it, if money is no object consider a HP/Dell/IBM …… server which is designed for full prodution and 24 x 7 running + SAS drives, your Xeons will run forever but your drives will fail first(moving parts rule)
Would you put budget tyres on a F1 and expect to finish a race.
Depends on access and application stack of course but a place I would start with unlimted budget (server only of course)
Hope this helps in some way.
Agree with Paul and I wish to add the following. For any database either new implementation or infrastructure refresh, some sort of capacity planning needs to be under taken to determine current cpu, memory, storage io/sec plus storage capacity GB needs to determined and allowance made for future growth. In your case your io subsystem really depends upon your io throughput required. If you do not have these numbers then you are really doing this by trial. Paul is right 0+1 will give you the best performance but is this io sufficient. You should also look at memory as well for example if you have excellent buffer cache hit ratio you will reduce stress on io subsystem. Also consider seperating data, index and redo logs on different spindles sets.
You will also want to look into separating different parts of the database onto different RAID arrays. Data should be seperate from Indexes which should be seperate from transaction logs, which should be separate from the tempdb and so on. The more you break things apart the better it will all work.