!! Please do not post Off Topic to this List !!
Hi Don,
I think I can get your "don't believe everything you
hear list" to 10....no make that 14. This is not
specific to any storage vendor :
8) I/O access patterns on datafiles and redo logfiles
are the same, hence can co-exist without any I/O
issues.
9) A logical device with 16 drives will perform
exactly like 4 logical devices with 4 drives each. So
always create one huge logical volume with all of your
drives.
- Even if you use Parallel Query and Database
Partitioning, you can still put everything on the same
large logical device. Don't worry about "localized I/O
isolation" it is not relevant.
- Don't worry about "availability issues" with the
one huge logical volume, even though you will affect
every database component with the failure of 1 disk
drive. That's because everything is mirrored.
- We have benchmarked this new I/O methodology on a
system with 1440 drives. We did another benchmark with
2400 drives and it worked really well.
- I/O diagnostics can be done only at the
file-level, as object-level I/O diagnostics is
extremely difficult if not impossible. Thus, create
one huge logical volume and put all of your database
components on the same logical devices to eliminate
hotspots.
- We are XXX Corporation, the gods of disks, and we
have invented an "7th fibonacci series inverse
convoluted extremely complex heat and pressure
sensitive algorithm", that will measure the angle of
the sun rays coming through the window in your
datacenter and take into consideration the time & date
of the day, determine the gravitational pull of the
moon, to manage the cache in our storage array which
will eliminate all I/O bottlenecks and cure cancer
automatically 7x24xforever.
Don't we love our jobs!!!!
Gaja
- Don Granaman <granaman_at_home.com> wrote:
> !! Please do not post Off Topic to this List !!
>
> WOW! Is the "other" workload on these similar when
> this job runs? Are you sure
> the problem is the Symmetrix and not something in
> the OS or instance
> configuration? Does this job spend a lot of time
> waiting (in Oracle) on physical
> I/O - or on something else? (I guess if you don't
> have access to the machine,
> you can't find out though. The ultimate tuning
> challenge!)
>
> If the problem is actually Symmetrix I/O, I could
> only hazard a guess that it
> might be due to RAID-5 for something inappropriate
> (hot redo log files?) or
> extreme I/O contention in the layout.
>
> As far as pointers, pitfalls, and suggestions... I
> really have only one: don't
> believe everything you hear!
>
> For example: (top 10 list)
> 1) "With EMC, RAID-5 won't matter." (it likely
> still will - for write-intensive
> stuff)
> 2) "With EMC, you don't want to stripe." (you might
> - it can still make a big
> difference)
> 3) "With the cache, I/O won't ever be a bottleneck."
> (until cache becomes
> saturated or ...)
> 4) [Corollary to #3] "Throw out all that basic I/O
> tuning stuff you learned"
> (but back it up to tape first!)
> 5) "The best layout is always SAME - stripe and
> mirror everything across
> everything."
> 6) "The check is in the mail."
> 7) "This won't hurt a bit."
> [ORA-00051]
>
> Drat! I crashed before getting to ten! Sorry, I
> was up all night repairing a
> bridge...
>
> For more serious and less evasive answers, see
> Gaja's paper at
> http://www.quest.com/whitepapers/Raid1.pdf
>
> -Don Granaman
> [OraSaurus - Honk if you remember UFI!]
>
> ----- Original Message -----
> To: "Multiple recipients of list ORACLE-L"
> <ORACLE-L_at_fatcity.com>
> Sent: Thursday, September 13, 2001 6:55 AM
>
>
> !! Please do not post Off Topic to this List !!
>
> Hi All,
>
>
> Does anybody here on the list have experience with
> EMC/symmetrix storage
> units.?
>
> We have our databases on this machine and I have a
> feeling the the I/O
> performance is not very good. I can not proof it
> since I do not have any
> experience/data/access to that machine. We do
> however have a very
> cooperative UNIX group but they also lack experience
> with performance on
> this machine.
>
> Who can give me pointers about I/O throughput that
> can be reached,
> configuration pittfalls etc..
>
> Example:
> RS6000 8CPU's and 4Gb memory with storage on
> EMC/symmetrix. Job takes about
> 2 hours to complete.
>
> F50 1 CPU 1Gb memory (TEST machine) local disks.
> same job takes 0.5 hours
> to complete.
>
>
> Jack
>
> =======================================================
Gaja Krishna Vaidyanatha
Director, Storage Management Products,
Quest Software, Inc.
Co-author - Oracle Performance Tuning 101
http://www.osborne.com/database_erp/0072131454/0072131454.shtml
Terrorist Attacks on U.S. - How can you help?
Donate cash, emergency relief information
http://dailynews.yahoo.com/fc/US/Emergency_Information/
--
Please see the official ORACLE-L FAQ: http://www.orafaq.com
--
Author: Gaja Krishna Vaidyanatha
INET: oraperfman_at_yahoo.com
Fat City Network Services -- (858) 538-5051 FAX: (858) 538-5051
San Diego, California -- Public Internet access / Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from). You may
also send the HELP command for other information (like subscribing).
Received on Fri Sep 14 2001 - 04:50:02 CDT
