[Date Prev][Date Next]
Re: New performance numbers
Le Jeu 11 octobre 2007 12:12, Howard Chu a écrit :
> Back in May 2006 we posted some authentication rate numbers for Symas CDS
> 3.3.2 (based on OpenLDAP 2.3.21.)
> On a 1 million entry database, we saw about 13K auths/second on a dual
> processor 2.8GHz Opteron system.
> We've recently re-run the tests using OpenLDAP 2.4.5 on a dual processor
> 2.8GHz dual-core Opteron system (so, 4 cores now instead of just 2) and
> 38K auths/second. One interesting thing to note is that we reached this
> rate on back-bdb, with the slapd server at 100% CPU utilization, as well
> with back-null, where slapd was only at 70% CPU. (This implies to me that
> there may have been a network bottleneck here, because we ought to have
> able to drive slapd into 100% CPU usage even with back-null. Even gigabit
> ethernet has its limits...)
> But the other nice point about this result is that while the number of CPU
> cores was increased by a factor of 2, the overall auth rates increased by
> factor of 2.7. So we can see that OpenLDAP 2.4 is at least 35% faster than
> OpenLDAP 2.3. It also shows that the lightweight dispatcher is still
> well with available CPU power.
What version of BerkeleyDB do you use? 4.2.52+patches, or one of the first
4.6 ? Do you use glibc malloc?
> Another point is the same thing I've harped on before - SLAMD is too
> inefficient on clients. We measured the 38K/sec rate using the slapd-auth
> client that I just added to CVS HEAD. Using the SLAMD java client we could
> only reach 29K/sec, even with 12 client machines and 10 threads per
> Someone's going to have to sit down and write a C client that speaks the
> server protocol and just fires off compiled C clients...
Do you know where the SLAMD server protocol is described? I (or someone
else, not very much time currently) could implement it in Charge (