Demo comments
Demo abstract deadline is (was) May 3?!
- We should have one or more ATLAS-specific demos as a top priority.
- Data management and replication using Magda
- Job submission using Grappa
- Distributed job management using Pavel+MOP+...
- Distributed grid-enabled hybrid event store
- Distributed analysis?
- pacman, packaging/distribution?
- We should also have a joint CMS-ATLAS demo.
- Baseline demo based on existing tools
- Demonstrate interoperability
- A real app from each experiment running seamlessly on the other
experiment's testbed
- Either testbed can accept jobs from either experiment using
site-specific policies
- Visual interface showing in a pleasing way the jobs running on
the testbeds
Schedule for joint demo:
May - demonstrate CMS simulation working on ATLAS
testbed and vice versa
June - develop policy scheme and implementation plan
July - meeting to evaluate progress and finalize demo
August - first version of visual demo tools
- initial infrastructure to run apps across domains (ivdgl-1)
See http://www.ivdgl.org/ivdgl-labs.html
September - first real ATLAS-CMS joint running
October - polish final demo
- interoperability with Europe DataTAG, EDG (ivgdl-2)
2001 demos
http://sc2001.fnal.gov/10.htm
http://people.cs.uchicago.edu/~voeckler/VDL/sc2001-demo.html
2001 demo notes
SC01 floor
(must be downloaded to be viewed properly in my experience)
From: Rob Gardner
To: Kaushik Default UTA ,
Torre Wenaus ,
"Jennifer M. Schopf" , Rich Baker ,
Bruce Gibbard , John Huth
Subject: [Fwd: SC02 research exhibit]
Date: Thu, 02 May 2002 09:28:04 -0500
We discussed some SC demo logistics yesterday - Randy Bramley
(CS prof here) wrote down some notes which might be helpful.
Rob
-------- Original Message --------
From: Randall Bramley
Subject: SC02 research exhibit
To: rwg@indiana.edu
CC: Shava Smallen
Attached is a preliminary view of last year's SC01 exhibit floor.
Note the vendors get the big splashy sites near the entrance; they
pay far, far more for their booths than research exhibits do.
If all you are planning on doing is a Griphyn/ivdgl sort of
demonstration and exhibit, consider just participating with the IU
booth at SC02. It will give you a chance to see what is involved,
and use the support that UITS provides. However, if this is to
be a major BNL thing, first contact people there who have done
this before. SLAC, LBL, ANL, PNNL all have exhibits there every
year, some of their people must have leaked over to BNL over the
years.
However, if you insist on doing an exhibit booth, here's all my
advice:
------------------------------------------------------------------
1) Decide early what size of booth you want to tackle. Beware that
a large booth is expensive. Sizes range from 10x10 up to 40x40 feet.
2) Assume nothing is provided, and you will be correct. In particular,
you must provide or buy/rent from the conference
a) carpeting if you want it
b) chairs, tables, table curtains, signs
c) network and electrical connections
d) lighting, or scrims to reduce lighting.
e) telephone, and telephone lines
3) The networking is typically not running reliably until the
evening of the gala opening, and sometimes not then. Buy a month
from some local ISP so you can at least dial up to a network
for the inbetween times.
4) The wireless networking worked well the past year, but this
varies. Do not let anyone run an ad hoc network in your booth,
it can earn you the death sentence (justifiably so) from the
SCINET people.
4.5) The SCINET people essentially put together in a few days
a major multi-gigabit network and handle a few hundred demanding
high-end users. They do work that would be worth a year's pay
at any company, and get zip for it. Be nice to them, even if
the network is crapping out. When there is a major outage or
problem, then they know it - and having 30 people insisting on
talking to them about how important their individual needs are,
only slows down the repair process.
5) Allocations are based on how long you have been continuously
displaying at SCxy conferences, within the size category for your
booth. Be prepared for a location that sucks. Also, if you plan
on doing this routinely in the future, select the size category
you plan on doing so that you are better placed in the preference
queue for future years.
6) Buy locally or bring along
a) spray cans of anti-static stuff advertised for clothes.
b) bottles of water for the local troops. Survival food
also helps, when they are trapped in the booth getting
something going overnight.
c) note that with costs of shipping and insurance, you may be
better off to buy linux boxes locally and then throw them
away after the conference than to ship them there and back.
d) lots of power strips, extension cords, tape for running
cords as needed, cabling connectors, hubs, etc.
7) Cut CD's with all the software you need or use, and all the
data sets. This will let you do installs, etc. while waiting for
SCINET to get running.
8) Typically, all the moving and electrical drops and some other
things (like carpet laying) have to be done by local union people.
Be prepared for major expenses in this category. Booze is not
allowed on the floor except for the gala opening, but a couple of
six packs hidden behind a table can make the workers much more
interested in getting your stuff done earlier.
9) Typically, you can start moving stuff in on Thursday night,
with the gala opening on Monday night.
a) Have someone get there as early as possible, so they can
direct carpet laying, moving in of boxes, etc.
b) They will kick everyone out around noon on Monday, so that
carpets can be cleaned, buffet food laid out, etc. However,
you can have someone stay at the booth *if* they remain
strictly in the booth, and don't wander the floor between
kick-out time and the opening.
c) You will mark your boxes with stickers, then get them back
to repack after the event. Tear-down takes one afternoon.
d) Decide early on about the manning of the booth, designate
a single booth-master and at all times have someone there
who can make decisions and answer questions. Have a
fixed schedule.
e) My personal advice: don't do demonstrations at fixed times.
Instead, contact the people who really need to see stuff
and arrange times when they will come see the demos. Then
have a sign-up sheet. You don't want a funding agency officer
coming over to see your cool stuff, only to find some student
decided that would be a cool time to be in the midst of a
Grand Theft Auto game.
10) Beware: if you are going to run a GRAPPA demo, it uses
Globus for authentication. You may need some fixed IP addresses,
so arrange to get those early on.
Last modified: Thu May 2 16:56:06 EDT 2002