Should I run my transaction log on Raid 1 or Raid 5? How about the data file
Right now I have them both running on Raid 5

Thanks for the help.

Ideally they should be on separate arrays and the Logs run best on RAID 1
vs. RAID 5. RAID 5 is better than 1 for the data but if you start to get
any performance issues disk related you may want to think about 0+1 instead.

--

Andrew J. Kelly
SQL Server MVP


"Eric Ward" <anonymous (AT) discussions (DOT) microsoft.com> wrote

Great thanks for the info

Another question: We are running Raid 5 and we just lost a disk (i.e. it went bad) and when the disk went out it corrupted our database file. Would Raid RAID 0+1 have prevented this corruption


----- Andrew J. Kelly wrote: ----

Ideally they should be on separate arrays and the Logs run best on RAID
vs. RAID 5. RAID 5 is better than 1 for the data but if you start to ge
any performance issues disk related you may want to think about 0+1 instead

--

Andrew J. Kell
SQL Server MV


"Eric Ward" <anonymous (AT) discussions (DOT) microsoft.com> wrote in messag
news:83787338-25DE-4437-A85A-97765E6BA851 (AT) microsoft (DOT) com..

If you have Raid 5 properly implemented you should be able to recover your database to the point it was before the one disk failure. Raid 5 will allow you to pop in a new disk in the place of the old disk and have the new disk rebuilt.

That's hard to say as it depends on what caused the outage and how it
occurred. RAID 0+1 is definitely more fault tolerant but there is always a
chance hardware can corrupt a file.

--

Andrew J. Kelly
SQL Server MVP


"Eric Ward" <anonymous (AT) discussions (DOT) microsoft.com> wrote

Raid RAID 0+1 have prevented this corruption?

RAID 5 has some bad overhead requirements for transactional updates. These
requirements increase in proportion to the number of drives in the RAID set.
Thus, the larger the spindle count, the poorer the transactional write
performance. RAID-10 performance is pretty linear for transactional writes
as the RAID set grows. For large writes, a good caching RAID controller can
limit the overhead of a RAID-5 array by 'saving up' the writes until the
entire stripe is ready to go to disk, thus avoiding the extra reads to
complete the stripe parity block. RAID-10 performance improves for large
block writes since the data is spread out over more spindles. These are
general rules, your particular controller or SAN may be tuned differently.

The entire point of RAID 5 or RAID 1+0 is to provide redundancy. If your
controller lost a single drive and corrupted the disk, I would consider it
either a malfunctioning or misconfigured piece of hardware. Either way, you
need to do something about it.

--
Geoff N. Hiten
Microsoft SQL Server MVP
Senior Database Administrator
Careerbuilder.com

I support the Professional Association for SQL Server
www.sqlpass.org

"Eric Ward" <anonymous (AT) discussions (DOT) microsoft.com> wrote

Raid RAID 0+1 have prevented this corruption?