RE: Oracle Performance on VMWare

From: Crisler, Jon <>
Date: Fri, 2 Jan 2009 12:57:12 -0500
Message-ID: <56211FD5795F8346A0719FEBC0DB067503945425_at_mds3aex08.USIEXCHANGE.COM>

I would be interested in reading about how you set up your benchmarks. Our benchmarks are a combination of home grown scripts and Quest's Benchmark Factory (trial version, still trying to get budget for a full license). The problem requires that you have at least 2 guests running on a ESX host , with the other guest idle (for best performance) and a benchmark consisting of many sessions, not just one. This is the most common real-world scenario. My thinking is that the context switches or thread switches are what is killing performance, along with somewhat worse I/O performance compared to native Red Hat. CPU intensive operations do not seem to be much affected, but apps that do heavy task switching seem to be much affected. We also tried hugepages with negative results.  

We use ESX 3.5, 64 bit RH4 or RH5 on HP servers with either AMD Opteron or Intel processors. We could very well be hitting some sort of bug or misconfiguration, but we have many engineers with many years of experience with ESX and have not identified the cause. An interesting note is that the problem affects both Linux and Windows, and affects Oracle, MySQL and Microsoft SQL Server 2000 / 2005, but does not affect other apps like webservers such as Apache / Tomcat, Websphere, Weblogic etc. I will take a look at your blog link, thanks for providing that info.  

From: Allen, Brandon [] Sent: Friday, January 02, 2009 12:42 PM
To: Crisler, Jon; Subject: RE: Oracle Performance on VMWare  

Hi Jon,  

Could you please provide more details on the bad performance you've seen with Oracle on VMWare? I've tested performance with 64-bit Oracle on OEL 4 & 5 both running on VMWare ESX 3.5 and performance was just as good on VMWare as it was on bare metal. I've also implemented a production environment in this configuration and another one running Oracle App Server 10g on VMWare with excellent performance as well. Maybe you were testing on an older version of VMWare, or VMWare Server instead of ESX, or had some misconfiguration or bug that was causing problems? For example, I had horrible performance when I ran 32-bit OEL4 on 64-bit VMWare, but when I switched to 64-bit OEL4, performance was great.  

Here is a good article about Oracle on VMWare (maybe a wee bit biased given the source):  

No, I don't work for VMWare, and I don't have extensive experience with it - only about 6 months, but so far, so good in my experience.  



[] On Behalf Of Crisler, Jon

Sent: Wednesday, December 31, 2008 11:18 AM Subject: RE: db_recovery_file_dest_size

Virtualization is pretty cool and has its place, but it has to be used wisely. High stress mission-critical environments that I would not hesitate to put on any other vendors hardware or OS, including RH Linux, IMO should never be put under VMware...performance is so bad. My findings are that production databases just don't belong on VMware unless the application is really small. You really hit the nail dead-on- its got to be low stress (i.e. light activity). Although performance is my big complaint, reliability seems to be fine.    

Privileged/Confidential Information may be contained in this message or attachments hereto. Please advise immediately if you or your employer do not consent to Internet email for messages of this kind. Opinions, conclusions and other information in this message that do not relate to the official business of this company shall be understood as neither given nor endorsed by it.

Received on Fri Jan 02 2009 - 11:57:12 CST

Original text of this message