The Leading Source for Global News and Information Covering the Ecosystem of High Productivity Computing
March 30, 2007
On April 22, 1977 General Telephone and Electronics sent the first live telephone traffic through fiber optics at the not-so-blazing speed of six megabits per second. Three decades later, optical communication is widespread. It's estimated that more than 80 percent of all the long-distance voice and data traffic is carried over optical fiber networks. Single fiber data rates are in the tens of gigabits per second.
But fiber bandwidth isn't everything. The bottleneck in optical communications comes when you have to convert the light into electrons. To do this today requires an array of discrete optical components. So while the cost of the optical fiber is cheap -- much cheaper than copper cable -- the cost of optical transceivers is not competitive with all-electronic solutions. The result is that at distances of 10 meters and under, optical communication is much less common than copper solutions. But as bandwidth requirements increase, the preference for optical interconnects over copper-based interconnects also increases, even at short distances.
Currently optical transceivers are built using a combination of expensive technologies. This includes discrete laser and photodetector components built from gallium arsenide or indium phosphide. These are connected to a printed circuit board along with dozens of other components and are interconnected by wires. The number of components that must be manufactured, assembled and tested makes current optical transceivers prohibitively expensive, except for long-haul applications.
Two recent announcements point the way to much less expensive optical solutions based on CMOS technology. Both IBM Research and Luxtera have demonstrated optical transceivers that take advantage of standard semiconductor technology to bring the core of optical communications onto silicon.
IBM researchers have built a prototype of an optical chip package which offers 160 Gbps of bandwidth -- 16 channels providing 10 Gbps each. This is eight times greater bandwidth than today's optical devices. The IBM optical transceiver has the driver and receiver circuits integrated onto a CMOS die. Other optical components, constructed from indium phosphide (InP) and gallium arsenide (GaAs), are added separately to the package, which is 3.25 by 5.25 millimeters in size. The optical transceiver is bump bonded in a flip chip assembly to attach the separate laser and photodiodes. The laser ends up sitting on top of the chip. The whole chipset uses just 2.5 watts of power.
The idea is to get the optics as close to the microprocessor as possible so that chip-to-chip communication can take advantage of the power savings and increased bandwidth offered by optical media.
Although IBM's press release mentions that the technology is capable of downloading a movie in a second, its first use will probably be in the data center. According to Marc Taubenblatt, senior manager for Optical Communications Group at IBM Research, this technology is especially interesting for HPC solutions, where the technology could be used to connect the microprocessor to I/O devices or the cluster interconnect fabric.
This will require printed circuit board manufacturers to add polymer waveguides to their boards and to develop compatible interfaces with IBM's optical circuitry. Mass manufacturing of optical printed circuit boards that incorporates the technology should enable low-cost optical solutions for node-to-node interconnects in a cluster.
"It's something that I think will take a few years for that ecosystem to become available in the market, but it's something we're actively working on," said Taubenblatt.
IBM presented the technology this week at the Optical Fiber Communication Conference & Exposition and the National Fiber Optic Engineers Conference, in Anaheim. The paper was titled "160 Gbit/s, 16-channel full-duplex, single-chip CMOS optical transceiver." What they described at OFC is the first generation of the technology. IBM has already prototyped the next generation in the lab that runs even faster. In this version each channel attains 12.5 Gbps reliably and can be pushed as high as 15 Gbps.
Page: 1 of 3(Digg, Technorati, more)
PGI Accelerator™ Fortran 95/03 and C99 compilers for x64+NVIDIA
Accelerate applications on x64+GPU platforms by adding OpenMP-like compiler directives to existing Fortran and C programs. Available now for Linux, MacOS and Windows. Download a free 15 day trial.
Platform HPC Workgroup Manager
Platform HPC Workgroup Manager integrates all the cluster productivity tools you need to deploy, run and manage your HPC environment.
Mar 19 | OfficialWire | New super to support intelligence work Down Under. Read more...
Mar 18 | ChannelWeb | Westmere parts already showing up in HPC machines. Read more...
Mar 17 | The Register | But what about the tier ones? Read more...
Mar 17 | Cadalyst Magazine | A new generation of workstations is changing the nature of technical computing. Read more...
Mar 17 | Linux Magazine | Latest iteration of Sun Grid Engine able to tap into Cloud. Read more...
Jan 12 | | In-depth look at vSMP Foundation server virtualization technology, technical implementation, use cases and capabilities. The technical whitepaper provides an architectural overview and details on the three vSMP Foundation products: vSMP Foundation for SMP, vSMP Foundation for Cluster and vSMP Foundation for Cloud.
Jan 18 | | This white paper discusses Gore’s copper cable assemblies, and how they continue to exceed the standards for providing reliable, cost-effective solutions for high-performance computer applications.
Join this online panel discussion for live Q&A with leading industry experts, analysts, and end-users to discuss the latest innovations, best practices, barriers to implementation, and measurable benefits of server virtualization with a particular focus on today's real world solutions.
Learn about scalable fault-tolerant architectures and examples of energy efficient and scalable supercomputing clusters using dual QDR InfiniBand to combine capacity computing with network failover capabilities with the help of programming languages such as MPI and a robust Linux cluster management package.
LIVE@SCO9: The IBM team discusses new innovations in hardware, software and services that help clients better understand their workloads and get insight from their R&D efforts. Technology demonstrations include the soon-to-be-released Power7 HPC processor, the DCS990 system with 2.4 petabytes of storage, the xCAT management tool, secure HPC cloud computing and more. Winners of two HPCwire Readers' and Editors’ Choice Awards! Take the IBM virtual tour at SC09 or more information go online to: http://www-03.ibm.com/systems/deepcomputing/sc09.html