[ghpn-wg] Suggestions for draft-ggf-ghpn-netserv-2
jukan at uiuc.edu
Wed Jun 22 13:23:38 CDT 2005
Michael, I have gone through all your comments and would like to
respond to some of them.
In general, your comments are going into the right direction - for
better organizing the structure and the layout of the document. That
was the main motivation for revising this draft. Below are some
> My question is: an "advanced network reservation" service
> is mentioned in some places in this document. Where is
> this service specified? Is this one of the more concrete
> things that was originally contained in netservices-0
> and was now removed? Then, where will it be contained - is
> a related document in the works?
This is open to discussion. In the current structure it is meant to be
under the Section 5 - "Service Profiles".
> # pg 5 pa 6, # pg 6 pa 1, 2, 3, 4
> ... all this is a repetition of previous text!
Do you mean Section 2.1? If so, do you mean it should be deleted,
shortened or rephrased?
This text does sound a bit introductory, but it is definitely different
from the intro itself. (or, I am missing the point.)
> pg 7 fig 3: I would change (b), as it doesn't reflect the horizontal
> nature of interactions at all - in fact, it looks quite vertical. There
> are no interactions between NS and grid services (or other NS's) shown.
Maybe the Figure does not reflect the "horizontal" nature of the
concept very well, but it is in my opinion needed as an alternative to
the "vertical" paradigm of Grid-network interactions. Think of it as
network running internally within the Grid service, e.g. such as
virtualizing a distributed machine (over network).
> My answer to the query outlined on pg 10 / 11 / 12: I would vote
> for option 4 (Franco's proposal). It seems like an "information
> service" would be something you query, and the information is not
> supposed to change all the time (things like the topology, and
> link capacities could go in there, I guess). I envision a monitoring
> service as an online tool that I can use to (for instance, manually)
> shift resources from one place to the other on-the-fly when I see
> that something goes wrong.
> Of course, the process could be automatic, but assuming about manual
> operation simplifies things in my opinion. If I were to manually run
> a Grid application, I'd first check the information service to guide
> my scheduling decision, then run it and keep checking the monitoring
> service to make updates while my application is active.
I like the idea of coupling one service to the application lifetime
(monitoring) and the other to the off-line analysis of the available
More information about the ghpn-wg