[OPEN-ILS-DEV] Hardware details

[Don McMorris]

Hi Bill:

In general, you can be _very_ flexible with Evergreen.  The OpenSRF
system on which Evergreen runs allows you to cluster very simply.

In the Blog archives was a "datacenter" tour for the PINES setup.
This may be a good place to look for some ideas.

Based on my memory, they have separate database servers and a cluster
of "worker" servers.  Load balancers distribute the HTTP traffic.
However, one design could have you putting the "external" interface
modules on some servers (the OPAC, reporter, and staff client
interface) while the other modules run on backend "worker" servers and
dedicated "Database" servers.

Another thing to keep in mind also is that systems can be reallocated
or expanded/reduced easily.  Say, for example, that you're finding one
of the workers' is causing some lag.  You can simply change some
configuration to load this worker onto other systems (or, perhaps,
change the configurations so that some of the servers dedicate more
resources to the worker).

I would tend to say the system design is limited only by your
imagination.  It is suggested that the OpenSRF system have a dedicated
jabber server system (as there are some configuration issues that need
to be made and may be undesirable for outside users), but beyond that
you can do what you want.

I think the "datacenter tour" in the OpenILS blog would give you a
pretty good idea of what has already been "proven".  But, again, it's
not the only configuration available!

Hope this helps! Please don't hesitate to contact us with any more
questions or request I clarify something I didn't communicate clearly!

Sincerley,
--Don

On 5/16/07, Bill Ott <bott at grpl.org> wrote:
> I see in the FAQ a description of the clustering, load balancing, etc.
> that you're using.  Seems like an excellent selection of tools by the way.
>
> I was wondering if someone could briefly describe the physical layout of
> this.  For instance, do you have a dozen servers acting as realservers
> all serving jabber, http, etc., then the database pool running
> separately, or is it abstracted further?  What are you using for storage?
>
> Budgeting for hardware here and wanting to consider various possibilities.
>
> -Bill
>
>
>