Re: avoiding corrupted block (ora-01578)

RAID5 is like any other tool; there are time where its use is appropriate & time where its use is less than optimal.

The benefits of RAID5 is that it does provide data level redundancy with "minimal" overhead is "wasted" disk space.

The specific performance problem with RAID5 is that it incurrs a one for one overhead write operation for ever data block being written. When a datablock gets written to a RAID5 volume the corresponding XOR block also needs to be updated or re-written. This additional write operation is needed to keep the redundancy data current so the RAID5 volume can recover in the event of a failure in one of the disks.

RAID5 can be beneficial where there is a high ratio of read operations in relation to write operations. So for tablespaces where the number of read operations is 10 or more to every write operation RAID5 will result in decent or improved throughput. On "read only" data RAID5 performs very close to using RAID0 (striping) because in both cases the data is spread across multiple spindles.

For example it would be a VERY POOR choice to place the archive log files on a RAID5 volume; because these files only get written to. Effectively the throughput would be reduced by about 50% to archive logs files on RAID5.

On Fri, 18 Jun 1999 08:25:46 +0100, "Jonathan Lewis" <jonathan_at_jlcomp.demon.co.uk> wrote:

> I'll second that very strongly.
>
>One of the biggest problems most installations
>have is that no-one sits down to review the
>technology and do the arithmetic, they just parrot
>an opinion that has been going the rounds for
>a couple of years and never pause to think
>that circumstances may change, technology
>may change, or that their situation is a special
>case.
>
>
>--
>
>Jonathan Lewis
>Yet another Oracle-related web site: www.jlcomp.demon.co.uk
>
>Jurij Modic wrote in message <376b8301.6512399_at_news.siol.net>...
>>On Thu, 17 Jun 1999 14:07:43 -0700, "Daniel A. Morgan"
>><dmorgan_at_exesolutions.com> wrote:
>>
>>>RAID 5 is a bad idea with any Oracle database under any conditions. If it
>does nothing else
>>>it will slow things down by a substantial factor.
>>
>>Sorry if this sounds a little rough, but the above statement is a
>>nonsense.
>>
>
>
>
Received on Fri Jun 18 1999 - 10:24:18 CDT