[Date Prev][Date Next] [Chronological] [Thread] [Top]

RE: SLOW Performance: OpenLDAP on Linux and IBM Dual Xeon?



Yes, my wa column of vmstat is up around 100 on my slow Xeon and only 9 or
10 on my kick ass Pentium 3. :)

Surprisingly, when I ran under single processor kernel, openldap seemed to
perform better, but I don't have the metrics saved to conclude that.  But I
can definitely say performance is abismal on my super fast machines and
superb on my jallopy.

> -----Original Message-----
> From: owner-openldap-software@OpenLDAP.org 
> [mailto:owner-openldap-software@OpenLDAP.org] On Behalf Of Donn Cave
> Sent: Wednesday, April 14, 2004 5:13 PM
> To: openldap-software@OpenLDAP.org
> Subject: Re: SLOW Performance: OpenLDAP on Linux and IBM Dual Xeon?
> 
> 
> On Wednesday, April 14, 2004, at 02:21 PM, Digant Kasundra wrote:
> > I am using OpenLDAP 2.2.8 with BDB 4.2.52, heimdal 0.6, and 
> cyrus-sasl 
> > 2.1.18.  All machines are on the same kernel with the same 
> cache size 
> > setups, same dataset, and the tests query the same uids.  The most 
> > bizarre part is that my Dell Dual Xeon 2.8 Ghz is only able 
> to do 5000 
> > queries a
> > mere 10 seconds faster than my Pentium 3 700Mhz.  And my 
> IBM Dual Xeon
> > 2.0Ghz comes in last, taking almost 3 times longer.
> >
> > I need to see if this is an OpenLDAP specific problem or a general
> > system
> > problem, as mentioned.  My HTTPD tests showed equal 
> performance and was
> > therefore inconclusive.  Is there another test suite I could try?
> 
> I had a host here that was a couple of orders of magnitude 
> slower than it ought to have been, oddly enough also a dual 
> Xeon.  In my case, all are new HP hardware, but only this one 
> is dual - though they're all running an smp kernel build.
> 
> I still don't known why, but it isn't a major priority since
> it never was intended to run a directory anyway.  The next 
> thing I am going to fix is syslog configuration - the fast 
> hosts route local4 network, the slow one to disk.  There are 
> a lot of little things that can be different, and somewhere 
> in there we may have a performance issue.
> 
> During your tests, is there a significant percentage of I/O 
> wait time?  (As "top" would report, or vmstat's "wa" column.)
> 
> 	Donn Cave, donn@u.washington.edu
>