Simone.
Where to begin?
Obviously you need a CE to accept jobs.
You need an SE in order to run any sensible jobs since
these jobs need to be able to work with datasets.
The UI is as you say the client tools for submitting
jobs and querying the other nodes directly.
WN (Worker Nodes) are the machines that execute the final jobs. For
instance
we have one CE that accepts jobs, these jobs are then passed onto
to a PBS[1] queue and then the job executes on one of the ten
worker nodes that we have.
So the worker nodes are in your local enviroments and could be
a 1000 machine farm , equally they could be one box. The CE gives
a grid point of access to what ever your worker node set is.
I cannot think of any reason why all these nodes' roles could not be
installed on one box. I would be interested myself if there
were any problems with this. It is something I have been thinking of
trying myself at some point.
Though if I could split them up to 3 boxes
a UI,
a SE,
a CE and WN combined.
I might. They can be really plain boxes, nothing fancy.
You can and probably should use other peoples RB, LB, ... for the
time being.
Globus does not have a RB. One has been written by datagrid that
does use elements of Condor. I think that's the case anyway?
Is your eventual aim to end up as part the EU datagrid?
Things worth reading
The Anatomy of the Grid: Enabling Scalable
Virtual Organizations on
http://www.globus.org/research/papers.html#Overview%20Papers
is almost compulsory if you have not already.
Some where there is a massive diagram of the EU datagrid which I cannot
find at the moment?
Hope this helps.
Steve
[1] http://www.openpbs.org/
-----Original Message-----
From: Simone Ludwig [mailto:[log in to unmask]]
Sent: Sunday, March 31, 2002 5:31 PM
To: [log in to unmask]
Subject: setting up a testbed
Hello,
I have a general question about setting up a testbed.
Is it correct that a testbed consists of the four Globus installation units,
CE, SE, UI and WN? (I don't know what a WN (worker node) is good for
though?) So, all the different units get installed on a seperate machine and
the CE, SE and WN can get accessed via the UI? Is that correct?
What about the other parts which are necessary for a Grid environment, the
RB (Resource Broker), IS (Information Services), RC (Replica Catalogue) and
the LB (Logging and Bookkeeping)? We need them too or? But I read that
GridPhyN is using Condor as a Resource Broker. Is there no RB in the Globus
toolkit then? What about the other parts (IS, RC, LB)?
Thanks a lot
Simone
----------------------------------------------------------------------------
--
Simone Ludwig
Department of Electronic & Computer Engineering
Brunel University
Uxbridge
Middlesex UB8 3PH
United Kingdom
Phone: +44 (0) 1895 274000 ext 2763
Mobile: +44 (0) 7900917544
e-mail: [log in to unmask]
Website: www.brunel.ac.uk
|