Re: Testbed

[Ian Foster <foster@mcs.anl.gov>]

I'd like to take issue with some points made in this note.

I believe that we will only make progress as a group if we start adopting 
common solutions.  Many groups have adopted GRAM as a protocol for 
accessing remote compute resources, GSI as a security solution, MDS-2 as an 
information/monitoring service, etc.  Now of course these elements of a 
Grid infrastructure are not perfect, and there is much working going on to 
improve them.  But the key thing is that by adopting these common elements, 
those groups can go off and work on other more interesting problems, rather 
than spending their time investigating interoperability.  Basically, GRAM 
(say) is an interoperability protocol.  So we can of course layer another 
interoperability layer on top of that, and another one on top of that, and 
... but what have we gained?  I really think this is a distraction.

To put things another way, Chip says "a homogeneous Globus GCE testbed does 
not test anything which is not already known to work" [and by definition is 
presumably not useful].  I think the whole point is that by adopting a 
homogeneous Globus GCE testbed, we can ignore a [relatively small and 
mundane] set of problems that have been solved, and go on to work on other 
more interesting and critical problems: like how to build schedulers, or do 
data replication, or how we represent policy, etc., etc.

-- Ian.

At 03:39 PM 6/13/2001 -0400, Chip Watson wrote:
>Gregor von Laszewski wrote:
>
> > 1) Java CoG can handle multiple credentials, a JSP interface is available
> > through JPDK, a Web interface based on CGI and HTML has been developed by
> > the Cactus group. Together with myProxy a convenient way to integrate this
> > in "dusty" web servers is possible.
>
>Similar solutions exist at Jlab.  Can they interoperate?
>
> > 2) I have to better understand what "running" Globus means. One way of
> > looking at this is Globus provides an Information service. A site may only
> > want to install this information service but not the rest of "Globus".  It
> > is easy to integrate other information services within this Information
> > services, called  MDS2. MDS2 has the security build in. If you do not want
> > to install MDS2 on your machine for one or the other reason, you have to
> > reimplement a service doing exactly this.
> >
> > 3) GSIFTP or GRIDFTP provide file staging abilities that can be 
> performed on
> > web servers through the Java CoG Kit. A rewrite of JavaFilesynchronizer
> > could do simple replication.
>
>Same point as above.  Can a system built from Java CoG play with a
>dissimilar system?  Are standards well enough defined?   (I suggest not, and
>that is one thing GGF should be working on.)
>
> > 4) We can technically submit remote jobs through web services.
> >
> > Thus, I assume Globus will be considered as part of the testbed?
>
>Yes, but a homogeneous Globus GCE testbed does not test anything which is
>not already known to work, so I propose we want to test multiple 
>implementations.
>(Although that is just my opinion).  It is at this point we must decide 
>what features
>are interesting to test.
>
> > Nevertheless, I also difficulties understanding (4) in your original mail.
> > Example: if I do not know what system or infrastructure I, I can not make
> > the decision what information/commands I should submit  to this
> > infrastructure. At least I must know my app is written in this
> > language/framework, it is compilable/runable on this platforms, I can send
> > this out to a broker that looks for resources matching my requirements.
>
>IMHO, a portal is more than a bunch of web pages (HTML) granting access
>to resources reachable by that web site.  It is also a bunch of web services
>which can be used to build applications at a higher level (as is being done
>in SOAP, ebXML, other industry initiatives).  What web services do we think
>are useful and re-usable?  That was the point of my list.
>
>Regards,
>
>Chip
>
>
>--
>Chip Watson
>High Performance Computing Group
>Thomas Jefferson National Accelerator Facility
>Tel: (757) 269-7101
>http://www.jlab.org/~watson

_______________________________________________________________
Ian 
Foster                                      http://www.mcs.anl.gov/~foster 

Math & Computer Science Div.            Dept of Computer Science
Argonne National Laboratory             The University of Chicago
Argonne, IL 60439, U.S.A.               Chicago, IL 60637, U.S.A.
630 252 4619 (fax 5986)                  773 702 3487 (fax 8487)