Email	Print Reprint
	add to:
	Del.icio.us Digg Google Furl	Slashdot Y! MyWeb Blink

InfiniBand is a communications link primarily used in high-performance computing. While its specification is rooted in technology from IBM, Intel, Mellanox, HP, Sun, Microsoft, and others, these days it is maintained and furthered by the InfiniBand Trade Association (IBTA). (For background on InfiniBand, see InfiniBand Technology by Corky Seeber.)

In 2004, the OpenIB Alliance, now known as the OpenFabrics Alliance, was formed to develop a standard Linux-based InfiniBand software stack. It subsequently shifted its focus in 2005 to support Windows in an attempt to create a cross-platform software stack. The goal of the OpenFabrics Alliance is to unify a cross-platform, transport-independent software stack for remote data memory access (RDMA), which is also designed to be interoperable with both InfiniBand and Ethernet. Then in September 2006, the IBTA and OpenFabrics Alliance joined forces to some degree to share advances in InfiniBand hardware and software advances.

InfiniBand features include Quality-of-Service (QoS) and failover, and is designed to be scalable. The InfiniBand architecture specification defines a connection between processor nodes and high-performance I/O nodes such as storage devices. Considering the demands being placed on bandwidth these days, technologies such as InfiniBand are becoming more and more important -- especially in the data center.

In an effort to find out what's going on in the world of InfiniBand, IBTA and market-research firm IDC recently conducted the first-ever worldwide forecast for InfiniBand adoption as a server interconnect for 2006-2011. The report, entitled Worldwide InfiniBand 2007-2011 Forecast and authored by analysts Stephen L. Josselyn and Lucinda Borovick, examines the technology for 2007-2011 and among other things finds that:

• HPC, scale-out database environments, shared and virtualized I/O, and increasing demand from financial applications with HPC-like characteristics have been driving and will drive more InfiniBand adoption.
• Factory revenue for InfiniBand host channel adapters will increase at a 29.3 percent CAGR from $62.3 million in 2006 to $224.7 million in 2011.
• Factory revenue for InfiniBand switches will increase at a 45.2 percent CAGR over the next five years from $94.9 million in 2006 to over $612 million by 2011.

The report goes on to say that while InfiniBand does seem to be getting traction in server clusters where high bandwidth and low latency are important. Moreover, blade servers, multi-core processors, and virtualization are putting more demands on datacenters, which is leading to migration to higher-speed, low-latency networking technologies, such as InfiniBand.

In short, the report seems to validates InfiniBand's growth and expansion in high-performance computing to the enterprise data center.

RELATED ARTICLES ZHMICRO Releases C++ IDE Integrating jQuery Client-Side Data Templates with WCF Encrypting the Internet Three 'Best' Persistent Model Patterns Announced Milepost GCC Now Available		TOP 5 ARTICLES No Top Articles.