IBM Expands High Performance Computing Capabilities in the Cloud
Product News Jul 31, 2014
IBM has announced that it is making high performance computing (HPC), as part of technical computing, more accessible through the cloud for clients grappling with big data and other computationally intensive activities. A new option from SoftLayer, an IBM Company, will provide industry-standard InfiniBand networking technology to connect SoftLayer bare metal servers. This will enable very high data throughput speeds between systems, allowing companies to move workloads traditionally associated with HPC, such as oil and exploration and data analytics to the cloud.
"As more and more companies migrate their toughest workloads to the cloud, they're now demanding that vendors provide high-speed networking performance to keep up," said SoftLayer CEO Lance Crosby. "Our InfiniBand support is helping to push the technological envelope while redefining how cloud computing can be used to solve complex business issues."
InfiniBand is an industry-standard networking architecture that delivers high transfer speeds — up to 56 Gbps — between compute nodes. That is the equivalent of transferring data from more than 30,000 Blu-ray discs in a single day. The architecture provides additional features, contributing to InfinBand's overall superior reliability, availability and serviceability over legacy PCI bus and other proprietary switch fabrics and I/O solutions.
This new HPC option enables very low latency between bare metal servers and private clusters of servers with up to hundreds of compute nodes, making it ideal for applications such as life sciences and genomics, computer aided engineering, financial services, electronics design and reservoir simulation. By reducing latency between bare metal servers in these private clusters, customers can easily manage massive amounts of data faster, more effectively and efficiently.
"Bringing InfiniBand capability to the cloud is driven by the growing need for extremely high levels of speed and performance for scenarios such as HPC and big data," said Philbert Shih, managing director for Structure Research. "This type of offering will help enable engineers and scientists to build, compute and analyze simulations in real time, leveraging hundreds of compute nodes. Being able to share and analyze data at this speed will only accelerate cloud adoption from this use case, while making HPC more accessible across a wide variety of industries."