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: Getting the best out of new server hardware: a disk setup for Oracle database

Re: Getting the best out of new server hardware: a disk setup for Oracle database

From: Mladen Gogala <gogala_at_sbcglobal.net>
Date: Tue, 13 Jun 2006 19:54:32 GMT
Message-ID: <448F1824.1020007@sbcglobal.net>


Heikki Siltala wrote:
> Mladen Gogala kirjoitti:

>> Well, let me re-iterate then: what do you need ASM for?

>
> + to get better performance: S.A.M.E should provide better performance
> than a failed attempt to distribute IO load evenly over the disks by
> manually relocating the datafiles

Any volume manager can do load balancing today. EMC has done it for years. Anyway, with your architecture, your bottleneck will be motherboard, not the disk array. PC servers are notoriously lousy at performing high volume I/O. With proper quantity of NVRAM on your disk array, you will start seeing bottleneck of the architecture around 1200 I/O requests per second. Linux box with 2 64bit CPU cards will not be able to sustain large quantities of I/O. That's precisely the difference between HP 9000 and Dell with 64 bit CPU's. ASM will not help you there. PC architecture lacks things like I2O, bus arbitration, separate cache coherency lines and data lines, proper write-back CPU caches and alike. To simplify things, on a PC-server, all I/O devices, CPU cache coherency, memory and video-card are on the same bus. The bus itself is extremely simple and doesn't provide any smart load balancing mechanisms like interrupt buffering, smart I/O arbitration or request piggybacking. When you stuff that box with 1200 I/O requests per second, network, virtual memory management and video card management, that bus will behave as any major highway with more traffic then it can handle. Despite the marketing hooray about Linux servers, one always gets what one pays for.

ASM is likely to make things even worse by requiring more memory, imposing yet another layer of software between SCSI drivers (which are in the middle layer on Linux platforms, and therefore slow) and the instance, requiring its own fair share of the fragile PC bus. I tend not to think in terms of marketing abbreviations but rather in terms of what those things really mean. A long, long time ago, the most popular measure of performance was MIPS, a marketing abbreviation that used to signify "Marketing Invention for Pushing Sales".

>
> + to minimize administration tasks: since there is just a huge ASM area
> based on raw devices and no datafiles I simply cannot waste my time in
> playing with datafiles :-)
>
> + to get best out of the licence fees: since we pay lot of money for
> Oracle licences we are willing to utilize all the features available -
> if not, we could move to MySQL or something! :-)

The wisdom, my young padawan, is to know which features to utilize and which to steer clear from. There are plenty of reasons for using Oracle, without ASM, which is not yet one of them. Oracle has outsourced development to Elbonia and has largely adopted the OSS model, which means that users are, essentially, beta testers. If you wish to entrust your data to ASM, don't say you weren't warned.

-- 
Mladen Gogala
http://www.mgogala.com
Received on Tue Jun 13 2006 - 14:54:32 CDT

Original text of this message

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