Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: ORACLE on Linux - IO bottleneck

Re: ORACLE on Linux - IO bottleneck

From: Mladen Gogala <gogala_at_sbcglobal.net>
Date: Thu, 09 Feb 2006 05:27:04 GMT
Message-ID: <pan.2006.02.09.05.27.04.39432@sbcglobal.net>


On Wed, 08 Feb 2006 18:05:51 -0800, Noons wrote:

> Having said that: we're running RHAS3-upd.5, 9.2.0.6, all patched up,
> and just direct IO. 8K block size. Without a saturated IO controller,
> I'm getting more than 500 direct IO/sec operations in a test program.
> But when the controller saturates, I get a lot less than that. It's a
> good
> indicator of when we need to start splitting things across
> controllers/hyperchannels.

Bear in mind that at some point all those interrupt requests will start saturating your system bus. PC buses, even the server versions, are nowhere near the capacity of the true midrange SMP servers like HP 9000 series or IBM P960 series machines. Modern disk controllers will do massive amounts of DMA communication with memory, using the same system bus that CPU boards use to synchronize caches, that network controllers use to notify CPU of the network interrupts and that all peripheral devices use to communicate with CPU and RAM. My experience tells me that no matter how good PC server you buy, you will never get more then 2500 I/O operations per second out of it. On a heavily used OLTP database, that amounts to 200-300 concurrent users, with up to 50 active at a time. When you get there, you are simply in need of a more powerful box. Linux itself makes thing hard to measure. One critical thing that Linux doesn't do for you is measuring I/O requests per process. The only monitoring utility that can do that is atop, with the home page at: http://www.atcomputing.nl/Tools/atop and it offers kernel patches that need to be installed in order for I/O accounting to work. Second, Linux doesn't show you the time spent on the interrupt stack. You cannot see whether your motherboard is loaded to the capacity or not, because you cannot see how much of the system time is actually spent servicing interrupts. On HP-UX GlancePlus does show you that. Linux is not an OS that I'd recommend for heavy duty serious processing. Linux kernel 2.6 is an unmitigated disaster and I would urge anybody to think twice before entrusting critical OLTP systems which must provide good response time to Linux.

-- 
http://www.mgogala.com
Received on Wed Feb 08 2006 - 23:27:04 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US