Grid Computing Economies Part 2 - Grid Media Store

Submitted by freelancejack on Mon, 2008-06-23 01:16

This is a possible use for the grid computing economy mentioned in part 1 , a media store funded by grid computing. The basic model works like this: a research group buys a large block of computer time for some major computational work they need done. That work is distributed in manageable blocks to users with spare processor cycles in exchange for credits.

These credits can then be exchanged for the right to download various media such as music, video, books and software. The server takes a cut and the rest goes to the author of the media.

This system is better for the users/crunchers/consumers than downloading the media for free. Users need to pay credits based on CPU time to download material however there would be no need for increasingly guerrilla advertising tactics to generate revenue from the download. The credits that the users pay are generated merely by having the computer on and not using the entire processor.

Because of the low distribution overhead, media could be priced in such a way that most users could download new material almost as fast as they can consume it. Say an hour-long music album costs $1.50 or 15 hours of credits. If the album is of decent quality, it can be listened to at least 10-15 times before it gets old and something new is desired, after which time a user would have another 15 hours of credits to spend. Likewise, if an e-book costs 10 credit hours (about $1), then it takes as much computer time to read as it does to buy.

Even disregarding copyright issues, a grid economy can beat a torrent system. Because the media is coming from a central server, users don't have to wait and hope for a seeder. In a torrent system, the average download speed cannot exceed the average user upload speed of about 50kb/s if people are generous. If a central server delivers the media through a torrent system like Blizzard Entertainment does with their upgrades, the average download speed can consistently exceed 200kb/s. So users still end up paying no money for as much media as they can consume while seeing no ads, but their downloads are four times as fast as a typical torrent. Therefore, for users, grid economies are better than downloading for free.

Returning to the album example, the creators of that album come out quite far ahead as well. For a $1.50/15 credit-hour album between $1.00 and $1.20 of that album can go to the artists after the costs of running servers for work distribution and uploading. This is about as much money as an artist receives if the album is sold in physical form at a retail store for $20, so the revenue per copy is the same. However, because people would rather pay no money for something than pay $20, they are likely to download a lot more music and a lot more copies will be 'sold' than through a retail store.

The media creators also gain a lot more exposure and will therefore be able to earn more revenue from future works or performances. As an added bonus, because it is more convenient for users to buy media with grid credits than to download it elsewhere without paying, fewer users will have the media that didn't generate revenue for the creators. Grid economies are better than either retail or free distribution for creators of media.

Finally, consider the people bankrolling all this media sharing, the buyers of the computer time. The examples above are based on the assumption that an hour of work on a desktop computer is worth 10 cents. This is the equivalent of about 35 cents per hour per dedicated CPU using the computers Sun Microsystems has for hire for large computations. However, Sun charges $1 per CPU hour. Because the cost of electricity, housing, coolant and maintenance is probably more than 35 cents per equivalent CPU hour buying time in bulk is unlikely to give time buyers a better deal than a grid system.

There's also the issue of availability. If a piece of work needs 200,000,000 desktop CPU hours, finding a cluster or group of clusters to do the work in less than a decade could be impossible. However, there are now hundreds of millions of desktop computers on the internet. If even a million users were willing to offer their spare computer time for free music and movies (consider seti@home has more than 12 million users (www.boincstats.com) that do this for no reward at all), then the work could be done in less than two months.

Some countries have implemented or are considering a tax on anything that causes carbon dioxide emissions, including energy hungry computer clusters. If a system is in place to prevent users from running their computers excessively when they wouldn't otherwise (like a ten hour per day limit on credits) then the commissioned computations would have almost no impact on carbon emissions and the tax could be greatly reduced. Grid computer economies are better than clusters for buyers of computer time.

*** Edit: Part 3 is available if you liked this story.

Bookmark/Search this post with: