Re: RAID 5 vs RAID 0

From: Zatherus <zatherus_at_theriver.com>
Date: Sat, 30 Dec 2000 16:54:28 -0700
Message-ID: <v9ts4tsneu24j5aiucpfq437vb044cuk6d@4ax.com>

On Sat, 30 Dec 2000 16:02:02 GMT, rugga_at_pacbell.net (Chuck ) wrote:

>
>
>I am trying to convince upper management to spend the bucks for Raid
>0+1 instead of Raid 5 and was hoping that someone could point me to
>some documentation of the web listing advantages/disadvantages between
>each. I know what they are but they want something more documented.
>
>Thanks greatly,
>CC

The Next text is right off of the OTN page. I suggest you sign up for OTN if you don't have it already. I believe it is still free and is an excelent Resource:
http://otn.oracle.com/files/search/search.htm?RAID%20white%20paper

Striping and RAID
Redundant arrays of inexpensive disks (RAID) configurations provide improved data reliability. However, I/O performance depends on which RAID configuration is implemented.

Below are the most widely used RAID configurations:

RAID 1: Provides good reliability and good read rates; however, writes may be costly.

RAID 0+1: Provides good reliability and better read and write performance than RAID 1.

RAID 5: Provides good reliability. Sequential reads benefit the most. Writes performance may suffer with RAID 5. This configuration is not recommended for heavy write applications.

Note:
Although RAID 0 provides the best read and write performance, it is not a true RAID system, because it does not allow for redundancy. Oracle recommends that you do not place production database files on RAID 0 systems.

Optimal stripe size is a function of three things:

Size of I/O requests to the array.

Concurrency of I/O requests to the array.

The physical stripe boundaries matching the block size boundaries.

Striping is a good tool for balancing I/O across two or more disks in an array. However, keep in mind the following techniques:

On high concurrency arrays, you must ensure that no single I/O request gets broken up into more than one physical I/O call. Failing to do this multiplies the number of physical I/O requests performed in your system, which in turn causes exponential degradation in your system I/O response times.

On low concurrency arrays, you must ensure that no single I/O visits any disk twice. To fail here causes the same performance penalty as the one described above. Received on Sat Dec 30 2000 - 17:54:28 CST