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

RE: 2.1 & 2.2 statistics, and some odd behavior that needs to be examined.



--On Thursday, February 12, 2004 4:16 PM -0800 Howard Chu <hyc@highlandsun.com> wrote:

-----Original Message-----
From: owner-openldap-devel@OpenLDAP.org
[mailto:owner-openldap-devel@OpenLDAP.org]On Behalf Of Quanah
Gibson-Mount


This explanation really does not explain what I'm seeing.  By
what you say
above, simple binds & SASL binds should see the same
performance issues,
because the memory pool will be getting dirty either way.
That is *not*
what is happening, if you read through my post on this:

SASL-based queries: 28 ans/sec average
anonymous queries: 222 ans/sec average

If what you are saying were true, I should have even *worse*
performance
with the anonymous queries, because they would be dirtying
the memory pool
faster.


That assumes that all else is equal, which it definitely is not.
SASL/GSSAPI Binds are inherently slower than Simple Binds, and the SASL
security layer adds encryption overhead to the protocol layer that is
absent from the Simple Bind case.

Yes, they are certainly slower, but they shouldn't be 1/3 of the speed of the same type of binds in 2.1. And still, if it is the dirtying of the cache that is so problematic, I would still expect to see the same type of issues with simple binds, which I don't. Whatever the reason, I don't see how anyone who uses SASL binds could even think of putting 2.2 into production the way it performs right now unless they are able to use the memory based caching.

--Quanah

--
Quanah Gibson-Mount
Principal Software Developer
ITSS/TSS/Computing Systems
ITSS/TSS/Infrastructure Operations
Stanford University
GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html