ACADEMIA

Breakthrough has broad future implications for computing, communications and embedded systems; initially enables compact, power-efficient and scalable parallel systems: In one of the most significant advancements in system-level technology in many years, Lightfleet Corporation has announced it has developed and is bringing to market a new type of interconnect technology, based on multi-channel broadcast light, to simultaneously and continuously link the key elements within computing, communications and embedded systems. This breakthrough all-to-all optical interconnect technology eliminates the internal complexity and congestion – mainly caused by the limited data-carrying ability and physical characteristics of the defined electrical pathways and buses that connect processors, memory, and storage. Congested communications are what have limited the overall performance of computing systems up to this point. Parallel processing systems using Lightfleet’s patented Corowave optical interconnect technology can, for the first time in history, operate in a truly parallel fashion, mimicking natural computing systems such as the human brain. Unlike traditional multiprocessor systems, a Corowave technology-based system spends much more time processing data, rather than waiting for data to be received from other parts of the system. As a result, system performance and efficiency are dramatically improved because all elements of the system have access to all of the information they need, at all times. Lightfleet developed and perfected Corowave technology over a four-year period. In August 2006, Lightfleet demonstrated a 16-node prototype as a proof of concept. Starting in July 2007, the company expects to ship its initial Corowave technology-based system, a high-performance enterprise-class server code-named Beacon. “Connectivity has replaced raw computation as the force behind addressing large problems and integrating the massive amounts of data being generated today,” said John Peers, Lightfleet Chairman and CEO. “In too many cases, people are facing impediments that should not exist. Lightfleet’s innovation can accelerate discovery and insight for business and science alike.” Benefits of all-to-all processing Today’s multiprocessor and clustered computers employ extremely fast processors connected by narrow and power-consuming miniature electrical pathways. Data must travel from point to point (serially) within the system along these pathways, a process that causes the system to drastically bog down under heavy workload. Although all of a system’s processors are operating simultaneously, they cannot exchange data with the same speed with which it is processed due to the system’s internal congestion. To eliminate the communications constrictions, the Corowave technology employs a straightforward array of optical transmitters, lenses, receivers and a mirror that allows all processors to communicate simultaneously. Data-carrying light is continually beamed and reflected internally throughout the system using these optical components. Because the light is spread broadly and reflected, and transmits at a very high speed, all light (and the data it carries) is available at all times to all connected elements in the system. Specific system elements automatically self-select and receive the data they require while ignoring non-needed data. The light-broadcasting system is compact and simple in design, and contains no moving parts. The compactness and lower power requirements allow extremely powerful computers to be designed with industry-leading form factors, and switchable power characteristics depending on computing needs at any point in time. Applications for cont inuous parallel computing True continuous parallel computing is perfectly suited for performance-intensive applications where large amounts of information are gathered and processed in real time. Systems using Lightfleet’s Corowave technology can broadcast and analyze multiple data streams simultaneously, speeding up applications such as data-mining, sophisticated pattern-matching and complex data analysis. Typical real-world pattern-matching applications include facial and voice recognition, fraud detection, cargo inspection, and real-time surveillance. Other relevant industry-specific applications include, but are not limited to, government (Homeland Security, military and defense, and aerospace); commercial enterprise computing (data mining, data warehousing, online transaction processing, voice response systems); medical imaging systems; and engineering, simulation-based and technical computing.