Re: ldap load measuring and reproduction tools

["J.Smith" <temp201@hotmail.com>]

Thanks for that insight - you actually are making a good point that I had
simply overlooked. Im almost ashamed of myself. ;)

But maybe I should have been a bit clearer about the exact kind of tests
that we are planning, so maybe I should explain a little bit more about the
reason behind the planned tests.

We are currently running a configuration with with 1 master server
and 1 single replica for our entire tree. For enhancing the availability, we
are planning to add a lot more replica's to the master. We will be
distributing all the ldap clients evenly over the readonly replica's. The
main idea behind this move is that more readonly replica's will improve the
availability of the directory.

The current plans are to add a total of about 10 to 20 ldap replicas to the
master.  However, this raised some questions. For example, there might be
a real world practical limit (or even theoretical limits) to the amount of
replica's you can add to a master server. Another concern is that the master
might be spending all of its time synchronizing the replica's, without being
able to answer any request/updates any more, or will at the least require
significant hardware upgrades.

So the primary thing we would like to measure and reproduce, is
modifications and updates/adds/deletes. Im not so sure that the exact
'search filters' still matter much in such a scenario, since we are
primairly interested in determining the scalability of the ldap replication
mechanism here ?


Sincerely,


John Smith.



----- Original Message ----- 
From: "Mike Jackson" <mj@sci.fi>
To: "J.Smith" <temp201@hotmail.com>
Sent: Sunday, March 13, 2005 9:28 AM
Subject: Re: ldap load measuring and reproduction tools


> J.Smith wrote:
> > Hi.
> >
> >
> > Im looking for tools that might be helpful with determining the amount of
> > load we are seeing on our live production ldap servers
> > (adds/deletes/modifications/etc), and that will allow us to reproduce a
> > similar kind of load in a test setup. Does anyone know of any tools that
> > will allow us to do such a thing, or that might make such a thing at
> > least a bit easier ?
>
> Hi,
>  Using simple operational statistics to reproduce load is not very
> accurate, unless you only use statistics from generic operations such as
> bind.
>
>  To reproduce a similar load in a test setup, you should analyze the
> logfiles from your production servers and record the search filters which
> were found, as well as the frequency with which they were sent. Use those
> filters with the appropriate frequencies to produce queries for your test
> setup. Basically, your job would be simple if you had a tool which could
> accept a logfile as input and produce replay traffic to the server of your
> choice.
>
>  I am not sure if such a tool already exists. If not, it would be fairly
> easy to write with perl.
>
> BR,
> --
> mike