Informix has advised us in the past to use the fragmentatiton facility in their version
7.1 rather than the striping capability of Online Disk Suite. I'm wonderingif anyone
has experience with Informix 7.1 and either GigaRAID/FT (or any hardware RAID
solutions) or Veritas Software's Volume Manager. Any comments will be
appreciated.
Tim Helton
ti...@amgen.com
We are using RAID 5 / Oracle V7.0.16 / Sparc Storage Array / Sparc Volume
Manager 2.0 / Sparc 1000
We have MAJOR performance problems, I/O bottlenecks, and are still trying
to determine the problem. We have applied so many patches that even our
patches have been patched. SUN has been unresponsive and we are
currently working with a consultant to figure this out. God help us!
Anybody want to buy a Sparc Storage Array? We know we can't get anything
for ORACLE!
--
Wally Beddoe Telekurs (North America) Inc.
wa...@tkna.com Stamford, Connecticut
Semper Fi! (203) 353-8100
Ever thought of switching to Informix?
I even heard from Informix about a way to convert Oracle Forms applications
to Informix 4GL.
Nils.My...@ccmail.telemax.no
NM-data, Dalsbergstien 7, N-0170 Oslo, Norway
My opinions are those of my company
If you use RAID 5 under a relational database you will, on average, get
performance problems, because RAID 5 performs miserably on writes.
I am suprised that SUN hasn't told you this. Before we bought the RAID
system, we asked HP, and they told us all the grim details about RAID 3 and
5, so we knew what we had to expect (we use RAID 5 under our Oracle database
anyway because we need the fault tolerance and possibility to hot-swap
disks).
To avoid the worst performance hits I have *very carefully* distributed the
database files among the three RAID-5 arrays using the logical volume
manager. Our server is no bolt of lightning, but it is adequate.
Using a RAID subsystem with a large battery-backed cache memory (or "RAID-7"
or other vendor-specific solutions) should improve the performance on
writes. HP's newer RAID systems have battery-backed cache, and our next buy
will probably be something like that.
Have you tried configuring the array to do striping, simple mirroring or
even individual disks (if that can be done with the SUN array?) instead of
RAID 5, and checked the performance?
--
+- Bo Forsberg -- Systems Engineer, ------------------- bo...@seisy.abb.se -+
! Information Technology, Computer and Network Services !
! ABB Industrial Systems AB dept. SEISY/QDD !
+ I speak only for myself (and to myself) but you are welcome to eavesdrop +
As for the write cache in a controller. We turned ours on just to see
how it worked. Its a good way to get depressed seeing how much time it
normally takes to do an I/O to a real disk as opposed to just stuffing it
in a buffer and saying its done. It made a big difference in our
performance, but our application is essentially write only so that is to
be expected.
Regards,
Jay Walters
Royal Hong Kong Jockey Club
Bo Forsberg (bfor...@ws262.sw.seisy.abb.se) wrote: