Last message on previous page:
In discussing virtual environments,

What is being described here is an embarrassingly parallel problem and is not really suited to the sorts of supercomputers I believe we have been discussing. This is really more of a distributed computing problem more than a supercomputing problem, along the lines of SETI or Folding @Home sorts of projects. While it is true that you will want a fast system to fully handle processing for some virtual object, and that this system will need to communicate state to another system, it's not something that would be solved by your typical supercomputer.

Hmm, ... putting a bit more thought into it, I suppose you could approach the problem this way ... multiple data, single program style. The singular application is only concerned about the physical properties and interaction of the object it has, based on metadata contained with-in that object, and the render of the object in the virtual space given nearest neighbor information on sources of light or shadow...

Interesting thought.

This pendulum has swung through here before.

In the 50s and 60s you could only interact with a system via a "remote" terminal. In the 70s, with the rise of the Internet you could truly be "remote," but still on a dumb-terminal. In the 80s with PCs, the clients get fat. In the 90s there was a resurgence of the thin-client (X terminals anyone?), but then clients got very nice and fat. Now there's a call for thinner clients again. Take a look at Sun's "Sun Ray" product for a modern example. Sun's reasoning is much more about the ecological impact of computing (power consumption) and the ease of the mobile worker, but it's still a wonderful concept.

This happens daily anytime someone interacts with a supercomputing system, though we have to watch that definition again. It's a CLI when talking to a true supercomputing system (at least today). What you're talking about as a "super-computing-server" is just a nice fat applications server. We run all sorts of applications like this on centrally hosted systems that are remotely accessed. In the land of X windows, this is trivial. For MS Windows apps we run Cytrix boxen.

Not true and not true, respectively. The whole point of doing this remote sort of work is that the data is on the central box you're working on, not on your local box. If the network blinks, you haven't lost anything but your connection. As soon as the connection comes back, or as soon as you move to another computer, you connect and find your work sitting there waiting for you just as it was. For three months I did fully graphical remote work from New Zealand back to the U.S. over a 36 kbps dial-up line five years ago. Since FreeNX wasn't up to speed (even around?) at the time, I was using a TightVNC server on my Linux desktop at work here in the States and had it setup to run a full Fluxbox "desktop" environment. I would use a tightvnc client, either from my Linux laptop or any Windows computer in an Internet kiosk that I could run tightvnc clients from. When I connected, my work was all there, whether it was a graphical installation process, a word processing document, mail in Thunderbird, web in Firefox, or whatever. It was all there.

As to not owning the software you pay for, ... I suggest you find another provider. Google offers free centrally hosted document applications. Everything I'm seeing and hearing out of the Microsoft side is that this is exactly where they are going (and bringing along with it the return of micro-payments). You'll connect to some MS Live server, do some word processing, and accrue a $0.03 charge for that hour you used their service. Add that charge up with all the other ones (spreadsheet, email, chat, or whatever they want to charge for) and you'll get your $10-$20 / month bill. This is a big reason why I support FOSS, I don't like, at all, the direction Microsoft is going.

Interesting you bring this up. You're absolutely right. In fact, I used my Treo 650 to do almost everything that I could do on a desktop: view, create, or edit MS documents; ssh into my systems; view PDFs; listen to music; watch movies; show digital pictures; play games. It was not a fully functional replacement for my desktop, but it sure was handy to have walking across campus or sitting in a boring meeting or walking through an airport.

Ummm, do you have you system connected to the Internet ever? You're certainly subject to someone owning your machine at anytime. Your response is going to be, "Yeah, but I use a firewall and (because you run Windows) virus protection and spyware detection and ..." I'm smart, but I trust the brains at Google a lot more than myself to handle security properly because I know those folks are like me, but there are more of them. Also, anytime you want your data, take it. Download it as a zip file or whatever you want and put it on a USB disk then walk away with it. It's your data, do what you want with it.

What is being described here is an embarrassingly parallel problem and is not really suited to the sorts of supercomputers I believe we have been discussing. This is really more of a distributed computing problem more than a supercomputing problem, along the lines of SETI or Folding @Home sorts of projects. While it is true that you will want a fast system to fully handle processing for some virtual object, and that this system will need to communicate state to another system, it's not something that would be solved by your typical supercomputer.

Hmm, ... putting a bit more thought into it, I suppose you could approach the problem this way ... multiple data, single program style. The singular application is only concerned about the physical properties and interaction of the object it has, based on metadata contained with-in that object, and the render of the object in the virtual space given nearest neighbor information on sources of light or shadow...

Interesting thought.


I was thinking multiple applications would be needed. One for the primary virtual object, as you descibe, and it's human interface/interaction. A second for the secondary and incidental objects and their relationship to the primary and user interface. Another for the environmental parameters and contstraints and it's interaction with the primary, secondary, and incidental objects, as well as the user. All of these applications would need to operate as a one.......in real time, of course.

Well, since I wasn't getting a straight answer from this forum, I did a little research on 'Grid Computing' and I begin to see why the super-crew doesn't like it. I don't like the idea of my super-computing-network reaching out to Other networks for resources, either. Nor do I like the idea of the processing process being unmonitored and uncontrolled.

I actually envisioned the system as being 'served' by a central computer, using only the resources on its local network. Each computer in the network might run a small process which somehow anticipates the difficulty of a given task and hands the bigger tasks off to the central server. The server then deals out portions of the task to local processors, retrieves and consolidates the results, and returns the result to the initializing computer.
Alternately, a User might know that he has a big job and send it to the server for super-processing. In a larger, more complex business, teams of super-servers might mobilize all of the computing assets of the entire business to do major processing.
Of course, the effectiveness of this model would depend on the speed and integration of the network-fabric. However, the key is that it is controlled by the network server and utilizes only the assets of the local network, however one wants to define 'local'.

Be Well!
Fireheart

I certainly think the model you're envisioning would be valuable to a lot of companies. Could there be a general framework set up where participating computers on a LAN run a background application? This application would provide a context for both donating and requesting distributed computing tasks.

I wouldn't want a "background application" running on my PC. Isn't that why we try to keep spyware and other crap from running down system resources? If it is voluntary, that's your choice, and it would have merits on corporate or university LAN/WAN's.

One "framework" is called Condor. Another is called Globus. Another is called BOINC.

Quite right, this is not something for the personal home computer... Unless you happen to be truly ambitious and have a home network that is usually on and a server to run it.

This whole discussion is about Business super-computing, not another iteration of Folding@Home, after all. And the process I envision running in the background would simply keep track of available resources and 'listen' for requests on the super-computing 'high-band'. It wouldn't have permission to interrupt processes initiated by the host computer.

Those are very interesting, Bropers. Too bad none of them are really ripe for general business deployment. Still, they're all works in progress and represent a foundation that could be built into a 'killer app'.

Be Well!
Fireheart

number crunching, weather models, Cracking a 128-bit encrytpion, lots of things.