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

Re: Benchmarks at Network World

On Fri, 19 May 2000, Julio [iso-8859-1] Sánchez Fernández wrote:

> Maybe you have seen this:
> http://www.networkworld.com/reviews/2000/0515rev2.html
> Anyone has any idea on what they used?  I presume it is the OpenLDAP
> as distributed by RedHat.  Presumably, it is compiled without threads
> as recommended for 1.2.x on Linux and I don't know what DB they used
> as backend, i.e. gdbm, Berkeley DB or something else.  We don't know
> the filters used in the test nor the indexing configuration.

- Unless there has been a drastic reengineering since the Umich3.3
days, slapd without threads would be a single process server. It
wouldn't even benefit from the OS level parallelism of the subprocess
forking model. Basically the code used a "thread" abstraction at the
server level that just ran the whole thing in a single process if
there was no support for threads in the underlying OS. It has been
quite a while since I looked at that part of the source, but I
remember thinking at the time that it would be very difficult to
modify it to use the standard unix fork/exec model of server.

> Anyway, I find some numbers shocking.  For instance, the messaging
> test with one client does 4.6 operations per second, while the same
> test with 10 clients does 47.7 operations per client.  I.e., it is
> scaling linearly with the number of clients.  Similarly, the
> addressing test did 5.9 operations per second with one client and
> 48.6 with 10 clients, i.e. going up with the number of clients
> but hitting some limits already.
> Any idea on what might be producing that?  Besides saturation on the
> client side, of course, that I will presume for now it is not the
> case.

- This is exactly what I would expect without native OS threading
support. Unless openldap has reengineered the slapd server to support
the standard unix fork/exec model, it's basically useless on a
platform that does not support native OS threads. I suspect Openldap
on linux is never going to win any speed contests due to linux's
thread model (threads are processes). However, I think openldap on
Solaris will be quite speedy. We did quite a bit of testing of the
Umich3.3 code before going into production and got really good results
(on the order of DNS lookups ie. 1000's per second). Jeff Hodges had
some web pages that summarized our testing, but like Jeff, they may be
long gone from Stanford's web site.

- Booker C. Bense