These kinds of scale GPU resources provides more flexibility to companies, particularly those customers who are performing research, engineering and development projects.
By offering high-performance computing in a pay as you go model, web hosting companies have a huge opportunity to reach a new market of customers.
The new HPC solution provides an on-demand, consumption-based model for those users looking to meet compute-intensive environments, such as more traditional HPC fields as data mining, numerical and seismic analysis as well as advertising agencies, game developers and interactive media firms with needs including video processing and 3D rendering.
“Up until now, GPU’s have been pretty much reserved for people that are willing to spend a ton of capital and have more data centers,” SoftLayer chief scientist Nathan Day said in a phone interview Tuesday. “We can make this available to the general market. It’s not quite the revolution that the dedicated server market started 10 years ago, but it’s a continued evolution of taking these unattainable processing units for a lot of small businesses and let them have it on the terms that regular SoftLayer customers are used to.”
GPUs work in conjunction with the server’s CPU to accelerate application and processing performance, offloading the compute-intensive portions of the application from the CPU to accelerate overall performance of the machine and significantly boosting overall application performance.
SoftLayer’s solution consists of dual-processor Intel E5-2600 (Sandy Bridge) based servers supporting one or two NVIDIA Tesla M2090 GPUs, delivering up to 665 gigaflops of double precision floating point performance.
As a result, SoftLayer users will see a massive performance boost with HPC applications accelerated by as much as ten times. SoftLayer’s HPC servers are available on-demand and can be provisioned via the company’s portal, mobile apps or API.
Some of the key features and benefits include a cost effective HPC which is ideal for scientific and technical computing, massive compute performance increase while accelerating application speeds by up to 10-fold, cluster management and GPU monitoring, parallel compute environment operates invisibly to the user, and GPUs specifically designed to work as computational accelerators.
SoftLayer HPC servers with NVIDIA GPUs start at $879 per-month for a configuration including one GPU, 16GB of RAM and 500GB of storage.
Last month, SoftLayer begin offering CloudLinux OS to its dedicated and cloud server customers, along with launching its new cloud storage service, SoftLayer Object Storage.
Talk Back: Are you currently deploying GPUs with your cloud servers. Would you say this is the general direction that the web hosting industry is moving toward? Let us know in the comments.