MAX IO Size on Tru64 5.1B

[Anand_31]

Hello,

I know this is a big topic of discussion but i am still taking a chance here!

how do i change the maximum size of an I/O request? is it a tunable kernel parameter or is it hardcoded into the kernel or derived from other parameters? this is from an Oracle database point of view, where we want to read the largest possible chunk (blocks) from the disk as allowed by the OS.

i remember reading somewhere about a maxiosize parameter..i didn not see such a parameter in the vfs or advfs subsystem on my 5.1B OS.

a search in this forum did not really get me the right answer...hence this post.

For ex. on Solaris, the tunable is called 'maxphys' and on HP-UX too i think we have a parameter that can be tuned.

we use Advfs and a 2-node cluster with an oracle database.

thanks for your help.
regards
anand

[Ralf Puchner]

as I remember there is such parameter but it depend on the used I/O system e.g. if LSM involved etc. So please post some detailed information about the disks used for oracle data (hardware, filesystem e.g. raw, lsm etc.)

[Anand_31]

Hi,

Thanks for that,

the SAN is an EMC symmetrix box and we use RAID 0+1. It is a GS1280 cluster running on 5.1B with PK3 with Oracle 9.2.0.4.0 database.

There is no LSM involved and i guess EMC provides the LSM at the storage box level and not on the OS. The filesystem is Advfs.

pls let me know if you need any more inputs.

thanks
anand

[Johan Brusche]

You could start with "sysconfig -q io", then read the manpage "sys_attrs_io".

Normally, it's the device driver that determines the I/O-size.

Johan.

[Anand_31]

Hi,

Thanks you both for the answers, i think we can find it.

regards
anand