Today Mellanox announced a collaboration with the University of Cambridge for the Square Kilometer Array (SKA) project. The University of Cambridge selected the company’s Virtual Protocol Interconnect (VPI) solution, consisting of ConnectX-3 adapter cards, SwitchX-2 based SX1036 36-port switches and cables, to provide it with leading interconnect performance and protocol flexibility for SKA test-bed clusters. The University of Cambridge and Mellanox will use the compute clusters for various development projects for the SKA project, an international effort to build the world’s largest radio telescope.
The computing requirements of the SKA exceed those of the fastest supercomputers, as the data processing and amounts of data generated by combining the signals from our antennas compete with that generated by the entire Internet,” said Dr. Paul Calleja, director HPCS Cambridge University. “Utilizing Mellanox’s protocol-flexible VPI solutions in the University of Cambridge supercomputer, we are able to run our applications over either Ethernet or InfiniBand on a single-wire to enable best application performance without the need for multiple networks or topologies.”
Delivering seamless data communication over FDR 56Gb/s InfiniBand and 40/56Gb/s Ethernet, Mellanox’s VPI allows customers to optimize their network for specific workloads and connectivity based on their interconnect fabric of choice across a common infrastructure. This flexibility is critical to allow customers to adapt to changing user and application requirements.
The SKA radio telescopes are capable of generating over 960 million Gigabytes of data a day. This enormous amount of data collection requires a highly scalable, performance interconnect infrastructure with flexibility to provide best application performance over any converged network protocol,” said Gilad Shainer, vice president of marketing at Mellanox Technologies. “With Mellanox’s industry- and performance-leading VPI solution, the University of Cambridge can provide the SKA with a computing infrastructure that can be specifically tuned to its advanced signal processing and computing needs.”
Read the Full Story.
Sign up for our insideHPC Newsletter.