Plura, a SETI-like Distributed Grid

Posted by Abel Avram on Dec 23, 2009

Sections

Architecture & Design,

Enterprise Architecture,

Operations & Infrastructure

Topics

Architecture ,

Deployment / Datacenter ,

Grid Computing

Plura Processing is a SETI-like distributed network harnessing the power of tens of thousands of computers.

Plura is a large distributed data center consisting of computer power contributed by users from all over the world. There are two ways to become part of Plura. Developers can embed Plura technology in their applications and, consequently, the users will run Plura code on their machine. The other way is for a website to include a Plura Java applet on its web pages, applet becoming active when the page is loaded by an user and it is performing very small computations on behalf of Plura. So, it is not the end user that becomes an affiliate of Plura, like in the case of SETI, but the intermediary, the application developer and the website. Both the application creator and the website need to fully disclose the embedded grid process and to request the end user’s assent for running the code. The Terms of Use (PDF) contains information regarding the entire process.

In both cases, Plura pays $2.6 for one entire month worth of processing code at 100% CPU usage. Considering that a website can have hundred of thousands or millions of visitors per month, each running small amounts of Plura code, it is easy to see how revenue can add up to significant amounts.

When an user accepts to run Plura code on his computer, the Plura applet downloads a small piece of a distributed algorithm and computes some values when computing power is available. In this way, highly intensive computational tasks can be run across tens of thousands of computers on behalf of customers.

One of the main concerns is privacy and security. Plura says they are not interfering with the host computer in any way by reading/writing/accessing information found in the computer. It just uses the CPU to run computations that are held in memory then sent over the network.

Some of the typical applications suitable for this distributed grid are: oil and gas exploration algorithms, financial modeling, bioinformatics, web crawling and Internet analysis, fluid dynamics, etc. Two such applications are Stanford's Folding@home project, aimed at analyzing and understanding complex diseases, and Malaria Control Project, tracking the spread of malaria. It is understandable why an end user would accept running such code on their spare CPU cycle. Other Plura-developer/website-user collaboration examples are: running code on behalf of a charity organization, running code while playing to get game incentives, or loading Plura web pages to support the development of a better website.

While the pricing is not disclosed, Plura advertises a cost that is 10 times smaller than running similar applications in the cloud.

• Older >