Big Compute
Windows Azure provides the best infrastructure for your Compute Intensive Applications
Big Compute applications require large amounts of compute power that typically run for many hours or days. Examples of Big Compute include modeling complex engineering problems, understanding financial risk, researching disease, simulating weather, transcoding media, or analyzing large data sets. Customers doing Big Compute are increasingly turning to the cloud to support a growing need for compute power which will provide greater flexibility and economy than having the work done all on-premises.
The new offering on Windows Azure provides powerful infrastructure with faster processors, more memory, and low-latency high bandwidth networking. This platform service is ideal for your High Performance Computing (HPC) applications as well as other compute intensive applications which require fast CPU’s, large memory configurations or very fast network connectivity between servers in order to solve a problem that are too big to fit in the memory of one machine.
Initially we will offer two configurations (instances) for Big Compute: 8 compute cores with 60 GB of RAM and 16 compute cores with 120 GB of RAM.
Both instance configurations are delivered on systems consisting of:
- Intel Sandy Bridge processors at 2.6 GHz
- DDR3 1600 MHz RAM
- 5 x 1 TB disks
- Two network connections
- 10 GigE network for storage and internet access
- RDMA + InfiniBand (IB) 40 Gbps network for inter node communication
Outstanding Performance and Efficiency for your Big Compute Application
The InfiniBand network in Windows Azure supports remote direct memory access (RDMA) communication between compute nodes. For applications, this allows memory on multiple computers to act as one.
Our RDMA network provides near bare metal performance with the elasticity and cost benefits of the cloud. Our testing shows latency of 2.1 microseconds to send a 4 byte packet across machines, and network bandwidth of over 3500 MB per sec. As a result, latency sensitive applications will scale better, with a faster time to result and lower cost. Windows Azure is the very first cloud provider to offer a virtualized InfiniBand RDMA network capability for MPI applications.
To demonstrate what HPC on Windows Azure is capable of, we ran the LINPACK benchmark and submitted outstanding results to the Top 500 List of the world’s largest supercomputers. We achieved 151.3 TFlops on 8064 cores with 90.2% efficiency. For you, this means better scaling for your applications with a faster time to result and lower cost.
Next Steps: