The parameters that go in deciding the raid depends on

• Criticality of the data
• Performance of the Storage
• No of Disks to be used for RAID Configuration
• Type of Hard Disks ( SAS or SATA )
• IPSAN or FC SAN
  

The usual decision usually centers around a RAID 5 or RAID 1+0 for a Block level data storage from Oracle servers

In a bid to understand the same, let us understand the basic of RAIDS

RAID 0 - This is where datafiles are striped across multiple disk volumes. But this is not true "RAID" since it offers no redundancy. If you lose a disk volume, you lose the data on that disk. RAID 0 does offer some nice increases in disk performance since multiple disk volumes are in use.

RAID 1 - This is where data on one disk volume is completely mirrored on another disk volume. If you lose a disk volume, then no problem. The system just gets the data from the other volume. The downside to RAID 1 is that you need twice as much disk. This isn't a terrible problem if your database is only 5 GB in size. But it can be very costly if you are building a multi-terabyte database.

RAID 0+1 (Sometimes called RAID 10) - This combines the best of the two above. Your datafiles are striped across multiple volumes and those volumes are all mirrored. You get very nice disk throughput and very good redundancy. Like RAID 1, you will need twice as much disk. If you can afford it, then RAID 0+1 is probably the "best" solution. Databases are typically write-intensive creatures. This configuration works well for database systems.

RAID 3 - This stripes data across multiple volumes. One volume is devoted to "parity bits". This bits are used to reconstruct data should you lose a data volume. Computing parity bits on write operations can slow those write operations down a bit. So database performance isn't nearly as good as RAID 0+1. But you don't need twice the disk either.

RAID 5 - This is similar to RAID 3, where parity bits are computed and used to reconstruct lost data. The difference is that the parity bits are not stored on a separate disk, they are striped across all disk, interspersed with the data. RAID 5's biggest advantage is that it uses the least amount of disk space for recovering lost data. So you don't have to buy as much disk. Unfortunately, RAID 5 has the worst write performance of all of these RAID levels. Write operations can be twice as slow compared to a normal file system.

No write intensive files should be placed on RAID 5 volumes otherwise you database performance can really slow to a crawl. People like RAID 5 because it is the cheapest solution. But the write penalties make this RAID level unacceptable for most Oracle databases.
This does not mean that Multi Tera byte databases don't use RAID 5 but the database gets updated once every two months. So the database writes are kept to a minimum. The database is "read-mostly". But any write intensive database files for this database (online redo logs, control files, etc.) should be +placed on RAID 0+1 volumes.

A clear winner in the IPSAN Space would definitely be RAID 1+0

Another way of looking at the above would be analysing the points in favour and not in favour of a particular RAID group

RAID 5 is better suited for mostly read-oriented applications. Because RAID 5 has to write parity information for each block of data it writes to the RAID set, it is not well suited for applications that are "write intensive" such as some transactional databases. It should also be noted that RAID 5 alone does not provide multiple-drive failure protection.

RAID 0+1 offers the same level of data protection as RAID 5 with the addition of multiple-drive failure protection. RAID 0+1 is also better suited for write intensive applications, as its mirroring component (RAID 1) does not need to calculate parity information when writing. In addition, its stripping component (RAID 0) offers the same read performance as RAID 5.

Now, looking at your applications: The MS Exchange Information Store is mostly read-oriented, so it is typically better off on a RAID 5. This is the most cost-efficient way to provide data protection while maintaining read performance. However, MS Exchange also uses logs that can get quite busy. These logs being mostly write-oriented, they will perform better on a RAID 0+1 array. You would have to evaluate the read/write ratio for your database application in order to determine the best type of RAID.

Of course, if your applications are not very busy, the performance distinction between RAID 0+1 and RAID 5 becomes much less noticeable and cost can become the driving factor. Ideally, you should consider a combination of both raid technologies if your disk array allows it.