• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Distributed computing company...

jonessoda

Golden Member
Let me just go ahead and say right now, that this concerns a for-profit enterprise, so if anybody decides that they don't want to advise me because of that, that's fine.

I understand that some businesses operate by selling usage for a distributed computing network; i. e. they have a series of networked computers and rent usage out to people who want to use them as a render farm, for math processing, what have you.

My question is, would this be feasible to do as a start-up business?

I suppose the initial expedenture and the overhead would be fairly large; around 100 low-end computers and a few high-end ones, a storage area of some sort that you can keep cool (probably an office location initially), networking hardware (gigabit switches/routers, a large-pipe incoming line), licenses for whatever for-pay software you use, salary for the technicians (I assume at least a staff of 5 or so would be necessary, all of whom would need to know how to keep it running, and at least 2 of whom are in the building at any given time that the system's operating)...

The market would probably be for, say, college students who need to render a graphics scene or animation and don't have access to a good render farm at their school (or access there is hard to get and takes a lot of time), for people really serious about contributing to distributed computing, smaller research teams who need to process data quickly and don't have their own DC network or supercomputer, etc.

Profits after overhead and salaries would presumably go primarily towards upgrading as necessary and adding new nodes, and then to advertising and other expenses.

Is there much of a market for this? Is success likely, or does a venture like this seem doomed to fail?
 
If I understand what you're describing, it sounds a bit like one element of "on-demand computing". Again, assuming that's what you're describing, then you're not alone - IBM and several other large companies are investing in this same market.
 
I don't know if it's "on-demand computing." As I understand it, ODC has a different concept, and the difference is one I'm not entirely certain how to put into words, but to sort of rephrase my idea (and it could be ODC and I'm just misunderstanding, that's a possibility), it's more like getting telescope time from a government to use a radio telescope. You pay for the usage, do your processing, and take your results away.
 
ODC is a fairly broad concept that can cover a whole host of applications. What I was specifically referring to was as you say, providing CPU power for specific applications as the need arises.

IBM and others are marketing toward bio-research and other firms that may need HUGE computational power for specific aspects of a part of a project, but don't want to have to invest large sums of money in the hardware to do so for a limited amount of time; with ODC they can lease time/processor power with a company that has the resources to provide it. Same basic concept to what you described: college students needing computational power for specific projects like rendering or animation, but don't have enough of those resources for themselves - and that's where you'd step in. 😉

 
It seems like IBM's (or Sun's or something) ODC inititive was failing, because nobody would pay what they were charging for CPU time.

Selling your own cycles to other people may not be profitable. In the past, cycles have had a very low market price. I have thought, though, that you might be able to consult setting up DC systems on the office machines of companies that want ODC.

Although that may be heresy around here, since many people have big herds on office machines. 😛
 
I think a big concern for companies using DC or even a company to do their work, is the sensitivity of the information/results. Tens or hundreds of millions of dollars could come from the processed results. Why would a company risk an information leak to possibly save 100-500k? Also, if IBM charged 10k to process, it would take less than 10 processes to buy 100k of equiptement. (tax write-offs and whatnot, i dont' care to do the exact computations right now)

I'm not sure if your idea of marketing to college studends would work out to well either. Just for the fact that they can get it done for free. It may take some time for them being in the processing queue but in the grand scheme of college students, one who would need an amount of processing power to warrent not completing it on their own machine would be very very small. (my guess less than 0.1 of a percent.) Also if the person really does need the results quickely, chances are, they'll be able to pull a string or two to get bumped up.
 
Back
Top