Sell CPU Cycles To CPU Share


Seti@Home may have stopped looking for ET, but don’t tell that to recently launched CPUShare, a market place in CPU cycles. Will it work? Not likely, even though I dearly wish for them to succeed, but I am such a fan of grids. It has a lot of shortcomings, the biggest problem is that it works on Linux, and there aren’t too many Linux desktops out there. The biggest concern I have is that the prices of hardware are declining so rapidly that companies or universities, the likely buyers of these CPU cycles can quickly build their own grids for a few million dollars. During the 1990s bubble many grid/distributed computing companies raised hundreds of millions of dollars in an attempt to build a processing marketplace. Almost all failed. I hope CPU Share avoids that fate, for it is a noble cause.

CPUShare has been founded with the goal of connecting together the computers of the Internet in order to create a general purpose Low Cost and World Wide Supercomputer available to everybody to use in a matter of minutes, controlled by a market for the CPU resources that chooses the price of the CPU resources using the supply and demand law in real time.


Jesse Kopelman

Since Sun sells there own Grid software and sells CPU cycles on their own Gird, I wouldn’t look to them to take the lead on fixing the hardware power.


While hardware costs are plummeting, most people overlook the fact that most of the IT costs are incurred in maintaining systems and keeping them updated with the latest patches, the right pipes etc. Also, they ignore the costs of maintaining a data-center or collocation.

Couple of things that will ensure that CPU Shares is successful are a good micropayment system for individuals willing to share out their desktops and a virtual machine.

John Powers

The existing desktops and under-used servers already available in most companies and universities are a great place to start a grid project. Most of these machines run a single operating system (Hint — it’s not Linux, not OS/X…), are linked by a decent network, and are already powered all or most of the time. When expanding the computing power available to an application, the step from one processor to thousands is far more valuable than the step from thousands to millions.


i think more interesting is if enterprises can, in effect, create their own internal CPU “marketplace” (and why stop at CPU? storage and other assets may also apply). Companies have servers all over the place dedicated to specific tasks, and which are under-used or have extended times of low activity. A company creating an internal marketplace, or slushfund, of IT assets can immediately tap in and take advantage of idle gear.

Om Malik

chauka…. here is the rub. the companies like Sun, intel and ibm are going to solve the power problem by building lower power chips and thuus th%re.

you can’t dig for oil when the prices go up but wait for more efficent engines from auto mmakers. same logic applies here as well.


I see my comment has already been made – electricity – there have been several articles about how the cost of power exceeds the cost of the hardware.

Luis Villa

IDC (I think it was IDC, maybe one of the other big analyst firms) believes that desktop Linux actually has a bigger installed base than OS/X, so it isn’t *that* small. If they can’t convince a few hundred thousand linux geeks to install a new toy, they are hosed anyway…



You are forgetting one VERY important detail. You have a logic error in your suggestion that the probability of CPUshare falling is likely due to the ability for universities to build their own at a fraction of the cost in the 1990s. Where is your logic error? Power. Many lately (Google, McNealy, Intel, et al) are focusing on the cost of powering these grids, not merely the cost of computing. Its the electrical grid that could end up being the huge cost (and the electrical grid is NOT a one-time sunk cost unlike hardware)!

Tsk Tsk Mr. Malik!

Comments are closed.