So what's the problem you're trying to solve?

If you simulate any database operation with OS utilities like dd you have to be 
sure you're simulating it the same way as database does. 

For example, Oracle opens its filesystem datafiles always with O_SYNC option 
and with O_DIRECT dependent on whether it's parameter filesystemio_options says 
to use direct IO. I'm not sure about how dd behaves. It'd be best to truss both 
processes and make sure same options are used.

If Oracle is slow doing queries then the DB people should first make sure 
whether their query actually uses an efficient execution plan before talking 
about IO. They could enable sql_trace for their queries and measure how much 
time these spend on waiting on IO.

Tanel.

-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Stephen Petersen
Sent: 10 June 2006 09:47
To: perf-discuss@opensolaris.org
Subject: [perf-discuss] Re: Re: maxphys and sd_max_xfer_size

Greetings,

I stumbled across this discussion and found it very interesting but I am still 
none the wiser to what is a better practice.

I found this when googling for poor I/O with a V440 and a 3510 much the same as 
mentioned in this discussion.  The V440 writes to the 3510 four times faster 
than it reads or so the database people say. To prove it, they used a similiar 
dd operation as discussed here but as the 3510 logical drives are setup for 32k 
blocks, they used 32k instead of 1024/8096k.  The dd wrote to vxfs filesystems, 
not raw devices.  Oracle is very slow when doing queries and the database 
people say the data in indexed and there are no joins for hell.  

I have not done this myself (yet) but if any of the settings discussed here are 
good or bad, I would love to know. 

I am assuming that Dave is the Ortera man and you know a bit about this.

Regards

Stephen
 
 
This message posted from opensolaris.org
_______________________________________________
perf-discuss mailing list
perf-discuss@opensolaris.org

_______________________________________________
perf-discuss mailing list
perf-discuss@opensolaris.org

Reply via email to