Cloud Computing Made Easy

Hi all, back after the holidays.  Happy New Year!

In recent posts, we’ve examined cloud computing platforms, infrastructure and software (all as-a-service).  Today, let’s touch on a couple other topics… high performance computing, web services, etc.

What are “Web services”?
The term “web services” is a bit oversimplified, and it implies that it is just a service that you access over the web. In reality, as a formal definition, web services are usually considered to be the domain of web programmers, not end users. It is a programming technique that involves use of remote subroutines, which can be called over the cloud, such s making a calculation or authenticating users. In the case of cloud computing, web services allow programmers creating cloud programs (SaaS) with ways to manage the cloud infrastructure, or integrate with other cloud programs. Using technologies such as SOAP, XML or WSDL, web services simply provide an ability to allow programmers to use other peoples’ offerings over the Internet.

Supercomputing-as-a-Service
Typically thought of as the domain of wild-eyed scientists working on large-scale projects that are far beyond the scope of ordinary business, supercomputing occupies a mysterious place in the computer business. But let’s draw a comparison—as recently as the 1970s, computing in general was thought to be the exclusive domain of a handful of extremely large companies and government agencies. Computers weren’t for ordinary people, or even for small companies. But look where we are today. The room-sized computers of the ’60s aren’t even as powerful as a simple netbook. Supercomputing today is in the same place that general computing was fifty years ago.
When supercomputing meets the cloud, then its power becomes available to a much broader audience. And that’s what is already happening. Maybe you can’t have a supercomputer in your home—at least not yet—but you can access one over the cloud. Companies like Exa sell their supercomputing processing power over the cloud, and companies that don’t necessarily have big budgets can harness the power of supercomputing environments. There are already a small number of companies that offer supercomputing as a cloud option, including the venerable Amazon, whose MapReduce offers supercomputer-like capabilities to crunch large data sets in Amazon Web Services.

High Performance Computing as-a-service (HPCaaS)
Along with supercomputing, cloud is also changing the face of high performance computing (HPC).

Supercomputing has always been expensive, often costing tens of millions of dollars. Nonetheless, they’re viewed as a necessary evil by many members of the scientific community. In recent years, grid computing has gained attention as a possible alternative. The notion with grids is to take advantage of otherwise idle CPU time that’s available on millions of computers. With grids, special software divvies up and “outsources” calculations to several computers in parallel, such as to PCs that act somewhat as mini-servers. Historically, grids were manually orchestrated and relied on other people’s computers, which raises questions about security and privacy.

With cloud infrastructure, we have already learned that servers can be allocated dynamically as needed (as in “thin provisioning”), rather than paying for unused computing power. Then, this begs the ten million dollar question: Why can’t I just harness the power of 100 servers when I need it, run a calculation and then shut them down? That way, I wouldn’t need to buy a supercomputer, right?

That’s exactly what HPC as-a-service does. Special HPC cloud software, including open source software like Univa UD, makes it possible to turn computing nodes on and off as needed, while orchestrating intensive calculations on those nodes. With cloud HPC, the concept is that a supercomputer never rests idle, doesn’t becomes comparatively outdated in a few years, and has no hard limits on scale. The future of cloud HPC is yet to be determined.

In the meantime, it will be fun to keep an eye on how the technologies mature for adoption by the serious scientific community.

________________________

Cary Landis, Virtual Global, Inc.

http://www.virtualglobal.com

http://www.teamhost.com

http://www.cloudipedia.com

Comments are closed.