startupmania

For the past several years, Google has made it its business what you can do with a few hundred thousand computers that work together and form a cumulative computing force.

It seems that the search giant has chosen to base its future on software developments rather than expensive hardware, and it believes that its unique software, that allows the hundred thousands of cheap computers to work together, offers better results than any commercial computing process currently offered by the high-performance computing (HPC) market.

Naturally, Google’s software infrastructure knows how to handle repeated malfunctions that occur daily in its vast network. Google also spends a fortune on building and maintaing the network that connects its dozens of data-centers together. Although Google does not reveal exactly how she does this, nor does it release a comparative analysis of its own system against others, it seems it does reach impressive results with its unique network.

If you’re interested by this topic, and would like to delve a little deeper into it – we recommend the various interviews with Google’s executives regarding their computer network, such as this one with Eric Schmidt. Another possibility would be starting from the Wikipedia entry about MapReduce, a framework which allows distributing a massive processing job among a large cluster of computers. For those who get really into it – check out the documentation for the open-source project Hadoop, you’ll be experts in no time.

After that lengthy introduction, it is clear that the concept of distributed computing, that has become so well identified with Google, or grid computing (as others call it), is no longer just a buzzword one reads in the latest article by UFO-hunting project SETI@home or other academic endeavors. It is a resource sharing technology well suited for massive data processing tasks, mainly for an organization that needs these type of operations in a non-sequential manner. Of course, this direction is fits very well with the shift of the computing world to the cloud, and works well with a cloud computing provider that offers a processing power to many external clients, such as Amazon.

That entire technological field is starting to get connected to the tech-buzzword “Cloud Processing”, and like any other new software field that is already assosciated with buzz, it needs to live up to the demands and capabilities touted by marketers and the decisions made by the organization big-wigs. Topics like reliability, security and, of course, ROI all need the appropriate amount of consideration.

One of the new Israeli projects in Cloud Processing, which tries to offer a data processing platform that will make use of the processing power of thousands of desktop computers not currently in use, and strives to offer an more commercially viable solution, is Green Cloud, founded by Zack Dvey-Aharon with the help of David Maman, which serves as the company’s director.

Green Cloud has already raised a seed investment from the Israeli venture capital fund Magma Venture Partners. The company employs seven development personnel in Hertzelia.

According to Dvey-Aharon, they already offer “a unique platform that is specifically made for cloud processing in organizational computers that are not in use during specific hours.” Their products is “equipped with the possibility to protect the local computer from running the code without the need for virtualization, as well as offer reliability technology that allows us to know that a calculation performer in a remote location is, indeed, an accurate calculation and not some kind of cheat.”

Dvey-Aharon promises that Green Clouds performs “smart utilization of hardware acceleration” and even offers crucial capabilities in the system management field, without hurting the user experience for the user who’s desktop’s processing power is “volunteered” to the general computing efforts.

“In working with Green Cloud, the customer only pays for the actual processing time, and not for ‘instances’ of virtual machines that work for you like in Amazon,” Dvey-Aharon explains their business model. “Likewise, we allow the capability to receive payments for agents that run on the customer’s machines and preform processing tasks. Meaning, a customer would be able to get paid for allowing his systems to be used by other organizations or private users.”

Dvey-Aharon adds that in the future the young start-up, composed entirely of former military-intelligence dischargees, “will change the perception of HPC,” he explains that “Today, many companies invest millions in servers in order to perform data processing takss, our vision is to revolutionize the costs of local processing. We already have a client for most desktop operating systems and in the future we will develop programs that will run on the HPCs themselves.”

Translated by Itai Rosenbaum