Hello from SC13 in Denver. We are excited to be here and talk with you about Big Compute and HPC at Microsoft. Our team is showing demos for Windows Azure and Windows Server with the HPC Pack, and we have partners including GreenButton and Violin Memory also showcasing their solutions in our booth.

Today we are excited to share that Microsoft is joining the steering committee of the InfiniBand® Trade Association (IBTA). The IBTA is a global organization well-known for guiding the InfiniBand™ specification, which provides high-throughput, low-latency communication links commonly used in high performance computing environments. The IBTA has also been involved in creating a specification called RDMA over Converged Ethernet (RoCE), which can achieve similar performances as InfiniBand but over Ethernet.

RDMA (Remote Direct Memory Access) networking enables one computer to place data in the memory of another computer with minimal use of precious CPU cycles, thus enabling very low networking latencies (microseconds) and very high bandwidths (over 40 Gbit/second) - all while using a negligible amount of CPU. To put this in perspective, using RDMA networking one can move the entire content of a typical DVD from the memory of one computer over the network to the memory of another computer in about one second, with almost no involvement from the processors of either computer

As an active member of the IBTA, Microsoft will help drive RDMA specifications and standards to enable performance gains and reduce networking overhead on the CPUs in large, mainstream datacenters. At the moment, Windows Azure has already adopted InfiniBand as the communication technology underpinning the hardware for Big Compute applications. In the future, we aim to bring cutting edge technologies like RoCE more broadly to Windows Azure.

With our RDMA-capable high performance virtual machines, Microsoft enables new classes of workloads to realize the scalability, elasticity, and economic benefits of the cloud. Customers can now leverage Windows Azure to accelerate discovery and insights from scientific modeling, including computational fluid dynamics and finite element analysis, with unprecedented agility, and performance that rivals first-rate on-premises clusters.

Alex Sutton
Group Program Manager
Windows Azure Big Compute