[AG-TECH] Question about AccessGrid small nodes

Gurcharan S. Khanna gurcharan.khanna at rit.edu
Sun May 10 23:09:55 CDT 2009


Yes, it was me. The larger project is the RIT Global Collaboration Grid, 
which envisions
using the highest possible technology to create a single virtual space 
for all of RIT's sites.

See http://rc.rit.edu/collaborationgrid.html for an overview.

The current architecture for this distributed yet coordinated set of 
nodes is simply standard
AG nodes. To reduce maintenance and enable others to contribute with 
minimum cost and
headaches, I have been asked to move closer to an appliance model, where 
I would specify
the hardware and software exactly so that it would be guaranteed to work 
first time, every
time. While I have been reluctant to go in this direction, it's 
obviously the only way to get
the buy-in I need to move quickly and credibly to deploy the 
collaboration grid. (Actually,
I would like to go even further and develop a 'thin-client' ag node but 
that's another story...).

So, we just ordered parts for ten little computers which balanced small 
form factor with cost
from NewEgg and GovConnection. Our original spec was to be able to send 
one DVTS
stream and receive at least five, using the extendedvideoservices in the 
access grid. We tested
the first one with this spec and it passed (except for the onboard GigE) 
so we ordered ten more
but with a PCI GigE card. However, we really probably want to do some 
H.264/Mpeg4 as well
but we need to run benchmarks for that still. Anticipating that we 
ordered five systems with a
dual core processor and five with a quad core processor to see what the 
difference would be.
The mpeg4 encoding has been quite intensive especially when you get up 
to 720 and 1080
resolutions so it's a real concern. We also required firewire ports for 
the DVTS. So far, we
have only run WindowsXP but intend to try out a linux as well. We are 
running AG 3.1 b/c
3.2beta breaks the mpeg4 and extended services that we need.

Here are the student-built computer specs:

Motherboard, AMD 780G, Phenom FX, MATX, Max 8GB $82.00
DDR2, PCIEX16, PCIEX, 2PCI, GBE, Vid, Aud, SATA,RAID Asus Computer Intl.

Processor, Phenom II X4 940 Black Edition 3GHz, 6MB L3 $240.00
Cache, 1800MHz FSB, Box AMD

ATX Desktop Chassis, PSI300 PS, 4U, 1394, Black $60.00
In-win Developement

1GB PC2-6400 240-pin DDR2 SDRAM DIMM $11.25 (x4)
Kingston Technology Corp.

Processor, Athlon 64 X2 5600+ 2.8GHz, 1MB L2 Cache, Box $73.00

Intel Pro 1000 / GT - OEM Single $27.00

80GB WD Cavier Blue SATA Internal HD 8MB Cache, $36.00
7200 RPM Western Digital

For audio (when needed) we are standardizing on Phoenix Duet Executive 
USB echo
cancelling mics at $180 and for cameras, the cheapest firewire Sony, 
Handycam® Camcorder $249.99 List Price. Output display devices will vary 
with location
and could be LCD panels or projectors. Different locations will also 
affect the audio solution.
We don't have an inexpensive solution for a good, single mic echo 
canceller for a non-tabletop

I'd be happy to elaborate on any of this and why we chose it and where 
we're going. It's
a somewhat moving target but we are trying to get something in place 
that will be adaptable
as well.


PS RIT Venue on the ANL server is where you can see where we're at at 
any point in time. -gsk

Purvis, Cameron wrote:
> Hello, from Cameron at the University of Alabama!
> I’m sorry if I have this mixed up: I believe that, in the AccessGrid 
> Town Hall meeting Tuesday morning, that I heard about RIT building out 
> small AccessGrid nodes based on inexpensive desktop computers. I 
> couldn’t make out the name of the speaker well but I think it may have 
> been you. The vision was to link several rooms together with 
> standardized units all in a grid venue. That sounds like something 
> that could be a very good demonstration of the technology here in 
> Tuscaloosa, Alabama. It also sounds like something that would be fun 
> to observe as the community starts to play with the technology.
> If you guys are working on something like that – a sort of 
> standardized design for small nodes – I’d love to be in touch with the 
> right person to talk about that. This is something I’d like to 
> replicate here, and would appreciate the chance to collaborate with RIT.
> Thanks!
> --
> Cameron Purvis
> University of Alabama Office of Information Technology, Research
> ckpurvis at ua.edu <mailto:ckpurvis at ua.edu> (205) 348-4677

Gurcharan S. Khanna, Ph.D.
Director of Research Computing
Office of the Vice President for Research

Assistant Research Professor, Ph.D. Program
Golisano College of Computing and Information Sciences

Director, Interactive Collaboration Environments Laboratory,
Center for the Advancing the Study of Cyberinfrastructure
Rochester Institute of Technology
1 Lomb Memorial Drive
Rochester, New York 14623-5603
Phone: 585-475-7504  ~  Cell: 585-451-8370
Email: gurcharan.khanna at rit.edu

More information about the ag-tech mailing list