Rescale, Intel, and R Systems have teamed up to provide an HPC cloud equipped with Intel’s “Knights Landing” Xeon Phi processors and the Omni-Path interconnect. The new platform, known as ScaleX Labs with Intel, is the first to bring this technology duo into the cloud computing realm.
The ScaleX Labs infrastructure is physically located at a datacenter in Champaign, Illinois managed by R Systems, a company that specializes in providing bare metal cloud hardware for HPC users. R Systems provides pre-configured clusters as well as more customized configurations for users that have their own mix of hardware in mind.
ScaleX itself is Rescale’s cloud management layer that sits atop the hardware and orchestrates the application workflow, file management, administration and other aspects cloud operation. In aggregate, Rescale provides access to over 30 datacenters and more than 120 HPC application packages – everything from ANSYS CFD and HMMER biosequencing applications to WRF weather simulation codes. Obviously not all of packages include Xeon Phi ports, but Rescale is promising some undefined number of “pre-configured and optimized codes” for Xeon Phi processors that can be tested now.
ScaleX Labs comes with four weeks of free computing to give users a taste of what is possible with the hardware. That free access is available with ScaleX Pro, the variant designed for small and medium-sized businesses. (Rescale also offers ScaleX Enterprise, a more full-featured version.) ScaleX Labs users will also get beta access to ScaleX Developer, the version designed for ISVs and other software developers who want to build ScaleX-capable applications.
The cloud offering seems to be designed to give more exposure to Intel’s Xeon Phi and Omni-Path products, which are in a fierce competition with NVIDIA Tesla GPUs and Mellanox’s InfiniBand portfolio, respectively. NVIDIA and Mellanox currently dominate their respective HPC markets and are widely available in both standalone cluster systems and high performance cloud set-ups. NVIDIA’s latest Pascal-generation GPUs, in particular, have been racking up wins in public clouds, the latest being Baidu’s addition of the Tesla P40 hardware for deep learning workloads.
That said, the four free weeks of Xeon Phi and Omni-Path could entice some HPC users to give Intel’s latest products a spin, while also providing a convenient platform for software developers looking to do port their codes to the new hardware.
According to our friends at HPCwire, renting a single Xeon Phi will run you $2 per processor/hour once the four-week free access period expires. No word on how many total Xeon Phi devices have been installed across the ScaleX Labs cloud.