Sunday, November 26, 2006

Warning: High Geek Alert (this post will contain much in the way of computers. If you are not interested, please stand by for a regularly scheduled post

Lately, Grid Computing has caught my interest. Mostly, I have been looking into the BOINC form of it. This is the software that runs Seti@HOME project, which uses computers around the world to search for a signal from an extra-terrestrial source. I can see the great potential in this kind of software but I am more interested in a commercial approach, which I will describe here.

There is one project called BURP, which renders 3D images and animations. Now, as you know, this is a similar proposition as rendering movies like Toy Story and other animated feature films. Normally, they use huge computer render farms to create the finished product. Now that home computers have become as powerful as they are. These kinds of computational tasks can be done by people at home rather than maintaining huge numbers of machines.

BOINC has a system of giving credit for each Work Unit (WU) completed. Normally people compete to be top of the credit list. This has led to some bizzare behavior for some participants but overall everyone seems to enjoy it. In a rendering aspect, think of a company or honestly anyone that wants to use a community like BURP to render their 3D cartoon masterpiece. They want to use the network but they have to either help out someone else's project (to build up credit) or buy credits from people on the network itself. Thus, credits become a commodity much like money. If there were a central place for people to join the market and parcel out WUs to members of the network, it would create a marketplace for people to buy and sell their computer time in an aggregate way. Those who allowed their computer to be used in this way more would reap the benefits of that work if someone needs to have their render done without contributing to the grid in the first place.

These credits could be bought and sold on a market exchange much like the stock market. People would place buy orders for, as an example, 100 credits at .01 cents (or pence) per credit. Once the requisite number of credits were purchased, the person who bought them could put a job on the grid without first having to help out on another project. Also, this would allow people to sell their computer time for a profit and recoup some of the cost of their computers in the first place.

There are of course a few problems with this vision. First, the bizzare behavior mentioned before would increase. People have tried all kinds of things to cheat the system to gain more credit. In the system described here, there would be even more incentive to cheat. Second, there would be even more incentive to try to figure out what was running on their computer. Kind of like collecting cells from a movie or something like that; a group of people could try to leak the contents of a film to a competitor. Third, how does a person build up this credit to sell on the market in the first place? This is a kind of chicken and egg problem. No project; no credit. People would have to volunteer their computers for a project first or there would have to be some way of gaining credit before you could sell it on the market.

Another large problem is the central authority for the system. Of course, there would be some kind of fee for every transaction on the market so that the system would run smoothly. How would this be regulated? A cooperative might be a way to do it so that everyone who registered would get some kind of credit. I am not sure.

This is all pie-in-the-sky kind of stuff. There are probably insurmountable problems (especially when you are handing off bit of someone else's information across the internet). What I do think will happen though is that we will see more and more of grid computing in our future.

posted by Chris  #3:49 PM | 0 comments |