When they’re not busy coming up with faster, lower-power processors for PCs, some of the best minds at Intel Corp. are trying to find ways to search more easily for enterprise information, map locations more precisely and network tiny wireless sensors that exchange data.That came out as scientists who are part of the chipmaker’s core research team provided an update on several works-in-progress during last month’s Intel Developer Forum.
While some of the projects have already seen deployments in customer environments, others are still in early-stage development, often in partnerships with local academic institutions.
Intel is working with Carnegie Mellon University, for example, on a search tool that will help users mine data that is not regularly indexed, such as photo collections, satellite or medical images. Code-named Diamond, the project aims to solve what Intel principal research scientist Rathul Sukthankar described as “the curse of cheap storage” — a growing inability to manage the reams of rich data in enterprise environments.
“We want to do for complex data what spreadsheets did for numbers,” Sukthankar said. “A lot of the data is easy to create or to store, but not to index.”
Most search technologies employ what Sukthankar called a “brute force” approach of sending a request to a storage device, which sends back every available result. “You end up throwing away 99 per cent of what you get back,” he said. The Diamond APIs, in contrast, would use machine learning and sophisticated algorithms to filter out and discard less relevant data more early. “If you’re searching for a needle in a haystack, you want to throw away the hay as quickly as you can.”
Intel believes Diamond could be applied in forensic labs to reconstruct information stored on video, or to pharmaceutical firms such as Merck, which is experimenting with the technology for possible use in its drug discovery research. Intel and Carnegie Mellon have already released early open source versions of a tool called SnapFind to search through collections of photos, and more releases will follow, Sukthankar said.
Mimicking radar
Stuart Golden is interested in finding things too, but his search project is focused on the physical world. As principal investigator of Intel’s precision location team, Golden is exploring technology that would have advantages similar to global positioning systems (GPS), but would also work well indoors and with greater accuracy.
Using a technique called time-of-arrival that was inspired by radar, an Intel-based device such as a laptop would send a message to a wireless access point which bounces back to the device.
Software would measure the time it takes for the signal’s journey and determine the distance it traveled. This information, which Intel hopes will be accurate within one metre, could then be communicated back to the network to provide location-based services.
“You could have an OS, for example, not only know where the closest printer is, but where the second-closest printer is, just in case the closest printer has a 200-page print job,” Golden explained.
Boston networks
Other potential areas for precision location would be in homes, where streaming video content could follow a user moving from one room to another, Golden said.
Intel’s more advanced research projects includes sensor networks, where tiny, self-contained, battery-powered computers with radio links called “motes” can organize themselves and exchange data. Dr. Ralph Kling, director of Intel’s Sensor Network Operation, said the chipmaker has already helped set up sensor networks in Boston using nodes based on its XScale technology to assist with water pipeline monitoring and in projects that would monitor the conditions of buildings during an earthquake. The XScale nodes allow the motes to travel through an 802.11 network more quickly, he said, boosting the performance.
“The Internet is mostly a virtual world. What’s missing is the physical connection to what’s happening,” said Kling, who predicted that sensors would convey information to bring the two worlds together. “It could be temperature, acceleration, position — all these things influence how business processes are completed.”
Intel sees market opportunities to use sensor networks to help customers in agriculture, retail and construction.
