We're gradually importing our scientific contents in French. Please forgive the visual glitches that may result from the operation.
IBTA Expands Interoperability with New InfiniBand Architecture Specification Release, Extends Virtualization Support
HPC Today  |  Wire  |  November 30, 2016
Note: the wired news below has been filtered but not edited by HPC Today.

New Vol. 2 Release 1.3.1 improves interoperability of the InfiniBand ecosystem, while the Virtualization Annex to Vol. 1 Release 1.3 enhances HPC and Cloud network scalability.

The InfiniBand Trade Association (IBTA), a global organization dedicated to maintaining and furthering the InfiniBand specification, today announced the public availability of the InfiniBand Architecture Specification Volume 2 Release 1.3.1 and a Virtualization Annex to the InfiniBand Architecture Specification Volume 1 Release 1.3. The new Volume 2 specification release expands interoperability and performance management functionality across the InfiniBand ecosystem for both high performance computing (HPC) and enterprise data center networks. The Virtualization Annex extends support for multiple virtualized endpoints within InfiniBand hardware. Both of these publications will be essential to expanding the performance and scalability of future data centers.

The Volume 2 Release 1.3.1 specification adds flexible low-latency Forward Error Correction (FEC) upgrades. This new option enables a best of both worlds – low error rate with low latency. A second enhancement enables the subnet manager to optimize both signal integrity with the lowest power possible from the port. Improved cable management is enabled through updates to QSFP28 and CXP28 memory mapping. The new specification release also improves upon InfiniBand interoperability and test methodologies for the latest data rates, namely EDR 100 Gb/s and FDR 56 Gb/s. This enhancement is achieved through updated EDR electrical requirements, amended testing methodology for EDR Limiting Active Cables, and FDR interoperability and test specification corrections.

To support the ever-increasing deployment of virtualized solutions in HPC and enterprise cloud networks, the IBTA also published a new Virtualization Annex to the Volume 1 Release 1.3 specification. The Virtualization Annex extends the InfiniBand specification to address multiple virtual machines connected to a single physical port. With this new annex, an InfiniBand subnet manager has a more efficient view of each logical endpoint, which significantly reduces the burden on the subnet managers as networks leverage virtualization for system scalability.

“As performance demands continue to evolve in both HPC and enterprise cloud applications, the IBTA saw an increasing need for new enhancements to InfiniBand’s network capabilities, support features and overall interoperability,” said Bill Magro, co-chair of the IBTA Technical Working Group. “Our two new InfiniBand Architecture Specification updates satisfy these demands by delivering interoperability and testing upgrades for EDR and FDR, flexible management capabilities for optimal low-latency and low-power functionality and virtualization support for better network scalability.”

The InfiniBand Architecture Specification Volume 2 Release 1.3.1 and Volume 1 Release 1.3 are available for public download here.

InfiniBand Continues to Lead HPC and Petascale Systems on the TOP500 List

The IBTA is also calling attention to the latest TOP500 List results that confirm InfiniBand’s standing as the leading interconnect family for the world’s most powerful HPC and Petascale systems. Currently, InfiniBand accelerates 65 percent of the HPC platforms on the list; which includes academic, governmental, and research sectors. For supercomputers capable of Petascale performance, InfiniBand represents nearly half of the segment at 46 percent, underscoring the continued preference of HPC architects to deploy InfiniBand given its unique combination of high performance capabilities.

Additionally, the IBTA is excited to announce the arrival of the first RDMA over Converged Ethernet (RoCE) enabled 100GbE system on the TOP500 List along with five 40GbE clusters. RoCE increases overall application performance for Ethernet-based systems through CPU offloads and higher utilization of compute resources.

Published twice a year and publicly available at www.top500.org, the TOP500 list ranks the world’s fastest supercomputer systems according to the LINPAC benchmark rating system.

Source: The InfiniBand Trade Association
With Business Wire

© HPC Today 2017 - All rights reserved.

Thank you for reading HPC Today.

Express poll

Do you use multi-screen
visualization technologies?

Industry news

S01-170208

Brands / Products index