When I was working on my thesis, I was faced with the problem of a lack of computing power for simulating the process under study. At that time, I got out of the kindness of the people around me, who allowed me to use two computer classes of OmSU for two summer months around the clock. Fortunately, the plan of my experiment allowed me to manually “distribute” the processed data on computers and everything ended quite well. However, the question of other variants of the distribution of computations got stuck in my head. And here I had an idea ...
I recently got acquainted with the basics of Adobe Flex technology. A fairly powerful, well-developed technology (offtopic: according to my observations, developing control panels is still more heroic than merit in Flex, although this use is often mentioned). Of course, not everything necessary is realized in it, but for my idea everything that is needed in Flex is there.
Also, I got to know the file hosting job closely. And he noted one, practically common feature among them - with a free download, asking the user to wait a few seconds. The seconds counter is "attached".
')
So I’ve got the idea of ​​a service “startup” that would:
1. Received applications from those who wish to distribute the processing of their calculations. Obviously, it is better to do this in the form of ready-made computational algorithms for ActionScript3.
2. conducted an additional preparatory revision of the “calculating” module, which will later be located on the web pages, contact the server for its “piece” of data and send the result of the calculation back to the server.
3. negotiated with file hosting sites or any other interested sites about the cost and order of placement of the “calculating” module in certain (logical for the user) sites of the site.
As a result, everyone should be satisfied: the visitor, having visited the site “paid” for some kind of necessary service, gave him the opportunity to “use” the computing power of his computer. The customer of the calculation receives the calculated data (it would be interesting to estimate how much teraflop such a “system” would show?) For a very reasonable price and time. The startup has earned almost nothing by doing :) quite a decent amount.
Of course, not all tasks are amenable to distribution. But there are such tasks. Otherwise, supercomputers would not be created or developed.
At the moment I have no opportunities and visible prospects to develop this idea independently. Therefore, I submit it to the discussion of habrasoobshchestva. It would be interesting to hear both the views on the proposed architecture and the ideas of the tasks that could be used for running the finished system.