web farm -newbie
Madison Kelly
linux-5ZoueyuiTZhBDgjK7y7TUQ at public.gmane.org
Fri May 30 19:09:56 UTC 2008
Jose wrote:
> Hi List,
>
> I am looking to learn how to setup a web farm, I have been reading
> reviews about some books like:
>
> Linux Enterprise Cluster by Karl Kopper
>
> High Performance Linux Clusters with OSCAR, Rocks, OpenMosix, and MPI
>
> Building Linux Clusters David H.M. Spector
>
> Linux High Performance Clusters [Import] (Paperback)
> by Alex Vrenios (Author)
>
> I got a couple of computer I rescue from the job dump and I would like
> to build a mini cluster, but not sure where to start, could you guys
> advice a newbie on this matters?
>
> thanks in advance
>
> J
There are many types of "clusters", so the first question you need to
answer is "what kind of cluster" do you want to build?
The cluster I have is a modest "HA" (high-availability) cluster. It
provides no performance benefit (in fact, it arguably slows things down
a touch), but it servers very fast "fail over" in the case of the master
node failing. This is probably the simplest cluster to build and can be
built on modest hardware. It's goal is "+9" reliability. That is, it
adds a '9' to your uptime.
There are N+1 clusters, which provides load balancing with hot spares
(very much like RAID 3/4/5). In this cluster, you have the load spread
out over X number of nodes who share, generally, a common data store,
with (at least) one more node than you need to maintain performance.
This is the logical progression from the first type, but requires
increasingly fancy hardware.
Web clusters generally use a load balancer that keeps track of multiple
web servers and assigns incoming requests to the give node with the
lowest load. These generally need to have their webserver nodes
connected to a common database, but this should be pretty simple given
that most RDBMS support network connections. This sounds like the type
of cluster you are interested in, given your email's subject. This
should be do-able relatively easily with commodity hardware. You may
even want to use two machines on a HA cluster for the load balancing and
backend DB and then a bunch of independant webserver nodes for handling
client requests.
Then there are render-farms and many, many other types of "farms" out
there. In the end, a "cluster" is much like a "server"; it defines a
type of service more than anything else.
To get started, take a look at http://linux-ha.org. It's were I went to
get started when I needed to build my cluster.
One last note; when it came to clustering the email, I had to roll my
own back-end schema for Postgres for my Postfix installation. Cluster
stuff is generally very individual, so exact implementations are
sometimes hard to come by.
Hope you have fun! I did. :)
Madi
--
The Toronto Linux Users Group. Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists
More information about the Legacy
mailing list