[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [AG-TECH] One-page summary of AG port usage -- please help uscomplete it

I'm gonna try to take a practical line here. First, some clarification:

- By default, the VenueServer allocates multicast addresses dynamically as follows:
address range: 224.2.128.0 - 224.2.255.254
port range: 49152-65535

- "Dynamic" in this case means that addresses are allocated when someone
enters the venue, and are recycled when the last person leaves the venue.

- "Static" (the opposite of Dynamic) means that addresses/ports are assigned
to a venue by the admin, and do not change.

- VenueServer admins can, through provided tools, configure the address
range however they like. There is currently no option for configuring the
port range, but there probably should be.

- The BridgeServer does provide a mechanism for configuring the port range
used for bridges. By default, it uses ports in the range 49152-65535.


One thing we could do is significantly narrow the range of addresses and ports
used by the VenueServer and BridgeServer by default. This would make the
default case easier for everyone to deal with. It would also, unfortunately, open
us up to more likely address redundancy across Venues.

Whatever is done, it should clearly be subject to the user community.

What are the concerns of your admins?
Do they care at all about opening up the multicast address range? Why?
Would it be enough to be able to select a bridge with a known port range that would work for any venue?

I think the real problem is in bridging, where you're concerned with traffic from
real hosts. But let's get input from some of those (Draconian) net admins and build
a practical solution.

Frank Sweetser, other net admins: Are you listening?


Tom Uram




On 2/13/06 1:05 PM, R. P. Channing ["Rick"] Rodgers wrote:
From: "Ivan R. Judson" <judson@mcs.anl.gov>
Subject: RE: [AG-TECH] One-page summary of AG port usage -- please help us 
complete it
Date: Mon, 13 Feb 2006 11:41:32 -0600


I think what you're hearing is unwillingness for anyone to commit to a wrong
answer.
From memory, the ports for vic, rat, and the rest of the tools are allocated
by the venue server. It can be a static allocation configured by the
provider or a dynamic allocation (I think the default). 
Perhaps my understanding is faulty, but I also believe that the default is
automated assignment by the server.  But that can not be entirely random --
the numbers must be drawn from a range of values appearing somewhere in the
code.  Having that range in a doc. is much better than having nothing,
which is what he have now.

A subsequent posting from Zsolt Nagykaldi makes the point, which I emphatically
agree with, that port assignment is a crucial problem to solve if AG is truly
going to take off.  He suggests using fixed port numbers with a registry, which
is one solution (and by "fixed" I believe he means that required ports would
remain the same for a given venue over time).  Another might be to adapt or
create a protocol allowing the exchange of port numbers when entering a venue
(which would at least tell you what ports were required, even if they don't
happen to be open at the moment).  Another might be to devise a tunnelling
mechanism through a number of fixed ports which are permanently assigned for
that purpose, sort of a port-related sibling to the multicast-unicast bridging
that the AG can do now.  This would allow people working in tightly constrained
networks to use the AG, at the cost of some performance penalty.

Whatever solution is ultimately pursued, dealing with the port issue is the
single most important single technical issue I see right now on with AG.  It
has certainly been delaying our deployment at the UCSSF medical center.
This document helps focus our attention on the problem, and I hope we can get
help completing it.  it needs to be accompanied by fuher instructions related
to port assignment (r.g., how they can be statically configured).

Cheers, Rick Rodgers


Given this configurability, it's impossible to state exactly what ports the
media tools will or won't use.

Sorry for the complicated answer.

--Ivan


-----Original Message-----
From: owner-ag-tech@mcs.anl.gov [mailto:owner-ag-tech@mcs.anl.gov] On
Behalf Of R. P. Channing ["Rick"] Rodgers
Sent: Monday, February 13, 2006 11:36 AM
To: rodgers@nlm.nih.gov; ag-tech@mcs.anl.gov; michael.daw@manchester.ac.uk
Cc: bernholdtde@ornl.gov; jsummers@bachman.cs.ou.edu;
terrazas@labmed2.ucsf.edu
Subject: RE: [AG-TECH] One-page summary of AG port usage -- please help us
complete it

Michael,

Thanks, I agree that AG Central would be an excellent development resource
and then final home for our summary of AG port usage.  We certainly need
help completing it.  The biggest gap right now is port ranges for vic and
rat.
I just *know* that folks working with the code for these applications more
actively than we do could produce these numbers in a few minutes.

In browsing around in the AG Central forums, it appears that they have
only
received light traffic thus far, not that that should stop us from helping
to
kick-start the community there.  It's not clear which forums would be
optimal
for our purposes -- I suppose "AG Toolkit/General"?  So I'm taking your
advice,
reattaching our two documents here, and also posting my original message
to
"AG Toolkit/General" on the forum at http://agcentral.org as wel.

CORRECTION: just tried to do the above, and got completely wedged -- not
clear that I have permissions to post to the forum, even though I'm
registered.
Then my web client got completely stuck.  Will try again later, sigh...

Cheers, Rick Rodgers


From: "Michael Daw" <michael.daw@manchester.ac.uk>
Subject: RE: [AG-TECH] One-page summary of AG port usage -- please help

us
complete it

Date: Mon, 13 Feb 2006 09:43:55 +0000

Great work guys! This is going to be good.

But, can I suggest using http://agcentral.org for this? If the document

is
discussed and refined in a forum, it will be easy to find. Once it's
ready, it
can be posted to the help center. Because most of us still aren't
completely
used to using agcentral, you can always post frequently to ag-tech too
with
pointers.



-----Original Message-----
From: owner-ag-tech@mcs.anl.gov
[mailto:owner-ag-tech@mcs.anl.gov] On Behalf Of R. P.
Channing ["Rick"] Rodgers
Sent: 10 February 2006 20:21
To: ag-tech@mcs.anl.gov
Cc: rodgers@nlm.nih.gov; bernholdtde@ornl.gov;
jsummers@bachman.cs.ou.edu; terrazas@labmed2.ucsf.edu
Subject: [AG-TECH] One-page summary of AG port usage --
please help us complete it

Dear AG Colleagues,

I now realize that the work I started last December, trying
to create a
one-page summary of AG port usage (based on the commendable
document created by
Javier Gomez Alonso of the Access Grid Support Centre at the
University
of Manchester) is not easily locatable in the list archives.
I resend it,
attached, along with the Excel version that David E.
Bernholdt of ORNL kindly
created.  As I said earlier, all of these documents are missing some
key information, such as the port ranges used by vic and rat.
 I send this out
again in the hope that another AG colleague will pick it up
and complete it.
We all really need to have something like this, and i would
hope that eventually
it would end up on the AG web site(s), and be maintained to
reflect any
coding changes/additions made to AG software.

Best Regards, Rick Rodgers

--------------------------------------------------------------
------------------
R. P. C. Rodgers, M.D. * rodgers@nlm.nih.gov * (301)435-3267
(voice, fax)
OHPCC, LHNCBC, U.S. National Library of Medicine, NIH
Bldg 38, Rm. B1N-30F2, 8600 Rockville Pike, Bethesda MD 20894 USA
http://lhc.nlm.nih.gov/staff/rodgers/rodgers.html






--------------------------------------------------------------------------
------
R. P. C. Rodgers, M.D. * rodgers@nlm.nih.gov * (301)435-3267 (voice, fax)
OHPCC, LHNCBC, U.S. National Library of Medicine, NIH
Bldg 38, Rm. B1N-30F2, 8600 Rockville Pike, Bethesda MD 20894 USA
http://lhc.nlm.nih.gov/staff/rodgers/rodgers.html



--------------------------------------------------------------------------------
R. P. C. Rodgers, M.D. * rodgers@nlm.nih.gov * (301)435-3267 (voice, fax)
OHPCC, LHNCBC, U.S. National Library of Medicine, NIH
Bldg 38, Rm. B1N-30F2, 8600 Rockville Pike, Bethesda MD 20894 USA
http://lhc.nlm.nih.gov/staff/rodgers/rodgers.html