Raw device vs. Fs. In HP-cluster

[Francesco_13]

Hi,
i have low performance with Raw device Rac Oracle database, the same database in FS run fine! . I have read many forum document in Metalink, but i don't know what change for eliminate low performance.
I use already ASYNC_IO = TRUE , i have a lot of buffer cache of OS..where i can find a solution?

Thanks.
Best regards.
Francesco

[Slawomir Gora]

Hi,

RAC is always slower than standard installation.
What is your hardware configuration ?
Do you use 1Gb link between cluster node1 for synchronization ?

[Francesco_13]

Hi,
yes i have interconnect with 1Gb and switch(private ip).
My hardware it's two node with 3 itanium processor for node and 10Gb of memory.
The same database in single instance in PA-RISC processor plataform (4 processor) it's faster.
The disk are the in the same box, in two plataform , HP-VA with san network fddi .

Best regards.
Francesco

[Slawomir Gora]

Hi,

I think that you should start with some performance test especially for disk: (glance, sar, iostat) - look at %busy do disks.

[Francesco_13]

Hi,
ok i can do it , but the Cluster it's free of any charge! I test with a simple large select in one only table in one only Oracle tablespace/raw device.
The wait for read it's high, but i don't know a reason. I/o statistics i think it's powerful for many access to all disks.

Regards.
Francesco

[Slawomir Gora]

Hi,

do you have the latest Oracle patches
- maby it is some Oracle bug.

[Francesco_13]

Hi,
yes i have lastest patch 9.2.0.5 RAC Itanium
Not bug with Raw in Itanium are reported in metalink..

Regards.
FRancesco

[Francesco_13]

Hi,
the problem it's in different speed when HP-UX use FS o RAw device. I will open other thread for this.
Thanks.
Best regards.