[Date Prev][Date Next]
RE: OpenLDAP, bdb and Linux filesystems
> -----Original Message-----
> From: owner-openldap-software@OpenLDAP.org
> [mailto:owner-openldap-software@OpenLDAP.org]On Behalf Of Quanah
> --On Saturday, November 22, 2003 5:23 PM +0100 Stephan Siano
> <firstname.lastname@example.org> wrote:
> > The first step (import) was to slapadd 10000 Objects (to a database
> > containing already 2 Objects). Five attributes were indexed in this
> > process. Since slapadd uses the same backend routines as the
> > bdb-backend, this should give some impression about write performance
> > (please correct me if I'm wrong here). I used a DB_CONFIG-file
> > containing only the line set_cachesize 0 134217728 1 (which results in a
> > cachesize of 160MB in one pool). A test without this file was aborted
> > due to my impatence.
> > The read tests showed that a sufficiently sized main memory will offset
> > all other factors. A fast CPU may further increase the throughput (but
> > the measured 650 search operations per second should suffice for most
> > real-life applications).
It would make sense that all the filesystems performed equivalently for
reads, since none of them have any journaling overhead in this case.
> An excellent summary, thank you!
> One quick question, did you have set_flags DB_TXN_NOSYNC in your DB_CONFIG
> file while doing the load? This can also affect load time (You want to
> comment it out after finishing the load and prior to starting slapd). I'd
> be curious to see how having that/not having that affected your load times
> as well.
He mentioned that his DB_CONFIG file contained only a single line, to set the
cache size. Using DB_TXN_NOSYNC will not have great benefit without also
bumping up the log buffer size; the default log buffer is small enough that
it would wind up needing to be flushed after most transactions anyway.
-- Howard Chu
Chief Architect, Symas Corp. Director, Highland Sun
Symas: Premier OpenSource Development and Support