By Matthew Dublin

On its official blog, Nvidia is saying that the next step for cloud computing is more GPU integration (no surprise there). While the majority of cloud computing is still dominated by Linux and Windows servers, for obvious reasons of functionality, there have indeed been more and more GPU cloud computing offerings sprouting up in the last year. Next to cloud computing, GPUs for scientific research has been one of the fastest growing areas of interest, so an intersection of the two technologies seemed inevitable.

But it's not the big cloud players like Amazon who have started to see around corner but rather, smaller outfits that are starting to include high-performance GPU technology as part of their service. Vendors like Penguin Computing was one of the first to get in on the act, with their high-performance GPU configurations integrated into the Penguin on demand (POD) cloud computing solution. Penguin has also started offering Nvidia's new high-performance based processing via "Realityserver," a software platform developed for handling compute intensive jobs. Nvidia used the roll out of the new platform at last year's Web 2.0 Summit to announce their intentions of entering the cloud computing space.

In July, PEER 1 Hosting launched the world's largest public GPGPU (general-purpose computation on graphics processing units) cloud service, which uses Nvidia cards and the Realityserver platform. And Sabalcore Computing, an on-demand high-performance computing on Linux servers, is also offering servers with Tesla GPUs and support for CUDA, Nividia's GPU programming language. Hoope Cloud, an Israel-based project designed to build cloud-based GPU computing systems based on Tesla GPUs, is still in alpha testing mode but already offers a variety of services including the execution of CUDA programs.

Does it sound too good to be true for the non-GPU/non-cloud computing research looking for a little HPC action? Despite continued improvements on the usability front for both GPUs and the cloud, neither is trivial to get up-and-running. But on-the-fly molecular modeling or protein folding, or imaging experiments, is certainly an interesting prospect and might, just might, make the double headache of the cloud and figuring out how to use graphics cards worth it.