[AG-TECH] Hardware news - towards a single machine node

[Allan Spale]

Ivan,

I think having AG APIs is a good idea.  Here are some of the APIs that I
think would be useful to include sometime in the future:

1) Abstract device driver API
Would allow for any device to have its functionalities integrated into the
AG.  I would assume data recording for Voyager should also be required.

2) Core API
Whatever APIs are necessary for the core functionalities of the AG.

3) Computational Grid API
This would allow people who would like to do collaborative high-computing
activities like visualizations to offload some of the processing power to
some other computing source designed for handling the tasks (as in the
case of high-quality rendering where some cluster would handle rendering
and then send the visualization information to the nodes...as I think I
remember seeing in some presentation during the AG retreat).  I am not
that knowledgeable about how AG will integrate into the computational
grid, but that is my thinking on that topic.

4) Accessibility API
Able-bodied people should not be the only ones who can use the AG.  This
would include support for participants with any disabilities like vision
impairments, hearing impairments, and lack of motor control to name a few.

5) Multilanguage APIs
AG is going global so there should be abilities to adapt AG stuff into
many languages or have translation services.

6) Mobile AG APIs
As bandwidth increases and technology improves, this will be more of an
option. That "super machine" that would do all AG stuff will eventually be
laptop-sized, so imagine having a portable AG node in a laptop.  Even more
so...an AG in a PDA.   

Just some thoughts for the future when designing stuff.  Although I think
that 1,2, and 4 should be integrated sooner than later.  5 would be nice
depending on how many non-English-speaking people are going to be using
the AG regularly.  


Allan Spale
EVL / UIC  node-op


On Sat, 31 Mar 2001, Ivan R. Judson wrote:

> 
> Hey all,
> 
> I'll be working in the next few weeks on a written definition of AG Node
> Requirements.  This will mostly be formalizing alot of the information we
> already have.  This will shake out what it is we expect the current 4
> machine node to do.  Defining what this functionality is, and building the
> API's for each component will let us explore this kind of idea without
> impacting the current operating schedule of the AG (ie, we're way too
> successful to have people randomly breaking nodes for experiments).
> 
> One part of this is to define "Core Node Functionality" and "Auxillary Node
> Functionality".  This is the line between what a minimum node is (the Core
> Functionality) and what a cool far out node might include (microscopes,
> milling machines, mood indicators, whatever).  If you have ideas for what
> you might augment a node with, please send those to me so I can include them
> as examples...
> 
> Any thoughts?  Ideas?  Are you going to toss me off the ship yet?
> 
> --Ivan
> 
> PS -- Oh my, we might actually start version numbering!
> 
> > -----Original Message-----
> > From: owner-ag-tech at mcs.anl.gov [mailto:owner-ag-tech at mcs.anl.gov]On
> > Behalf Of Jeff Ramacher
> > Sent: Friday, March 30, 2001 7:15 PM
> > To: Todd Needham
> > Cc: disz at mcs.anl.gov; Ag-Tech at Mcs. Gov
> > Subject: Re: [AG-TECH] Hardware news - towards a single machine node
> >
> >
> > I think it would be really cool (but probably way too expensive to
> > develop) to move some of the AV processing off of the system CPU to a
> > special purpose AG pci board or something like that.  Does anyone know
> > if there's a "vic in a chip" kind of thing already?
> >
> > Just my thoughts on a Friday afternoon...
> > --
> > jeffr
> >
> > Todd Needham wrote:
> > >
> > > Given discussions regarding CPU requirements for a single machine node,
> > > thought you might like to know I saw a pre-production dual P4 (1.7GHz
> > > Foster) from Dell today: 64 & 32 bit PCI buses, built-in 1394, onboard
> > > fast Ethernet.  The real news is that they expect to start shipping
> > > these in May with a reasonable starting configuration under $4K.
> > > Needless to say, 4 cameras and all those video streams are not going to
> > > be an issue.
> > >
> > >
> > >
> > > We also had discussions with Viewcast/Osprey about getting P4-optimized
> > > drivers for their capture cards and we're following up separately with
> > > their development team next week.  It would still be nice to get all of
> > > this running on a single proc P4 to keep the cost down even further.
> > >
> > >
> > >
> > > Todd Needham
> > >
> > > Manager, Research Programs Group
> > >
> > > Microsoft Research
> > >
> > > http://research.microsoft.com
> > >
> > >
> > >
> > >
> >
> > --
> > Jeff Ramacher (jeffr at mhpcc.edu)                | (808) 879-5077, x264
> > Maui High Performance Computing Center (MHPCC) | fax (808) 879-5018
> >                                                | http://www.mhpcc.edu
> >
>