Re: Road Map : improved scalability - how to proceed ?

["Medievalist" <ldap@HBCS.Org>]

On the subject of scalability:  I'm kind of surprised by this, but I'm the only 
OpenLDAP user I know who likes to run lots and lots of replicas.

I prefer to put a replica on every server, just like I put a DNS slave on every 
server... and just like with DNS, queries into the service are restricted to 
the loopback address (I probably should be using ldapi: instead of loopback but 
I haven't created time to experiment with that yet).

Master servers only exist to feed replicas, be backed up, and to provide a 
single place for updates to be entered.

I will be happy to provide feedback on how many replicas I'm successfully 
running at any given point, but if somebody (with the resources) can figure out 
how to test a truly enormous number of replicas (say, several thousand or 
hundred thousand) it might reveal some interesting limits to OpenLDAP's 
scalability.

Just a thought.  No need to attach any importance to my maunderings.

--Charlie

On 12 Apr 2004 at 16:29, J.Smith wrote:
> 
> I just browsed through the openldap roadmap, and noticed that improved
> scalability and/or performance is coming up on the horizon sometime soon now
> (~3Q2004 ;). I was just wondering, how should we start looking at measuring
> and/or improving the scalability of OpenLDAP ? Should we start creating our
> own (mini-)benchmarks and scalability tests ? Or should we simply start out
> by using something like Mindcraft's DirectoryMark, and see where that leads
> us ? Should we do autobuilds regularly, followed by a testbench run and
> autopost the results somewhere to see how good we are doing ?
> 
> Anyways, any and all ideas are more than welcome here.
> 
> 
> Sincerely,
> 
> John Smith