A neonatal intensive care unit. A buoy tethered deep in the waters of Galway Bay. A space center in Sweden. All three are sites where stream computing is being tested as a powerful new way of processing data. IBM's new middleware platform, also known as InfoSphere Streams, can ingest and analyze massive amounts of diverse data in real time and issue predictive bits of intelligence that can help its users make smarter decisions... about caring for critically ill preemies... managing a fragile marine ecosystem... and forecasting disturbances in "space" weather.

Plugging into the real world
Stream computing surfaced several years ago at an IBM Global Technology Outlook conference, a meeting where scientists put forth their most promising ideas. "In simple terms, instead of plugging a computer into another computer, the premise of stream computing is to attach a computer to the real world—to people, to traffic, to the environment—and see what happens," explains Nagui Halim, the researcher who leads the stream computing initiative. "Once you collect the data, you have to analyze and visualize it... to understand what is going on. This is the concept of 'real world awareness', and it turned out to be one of the underlying principles of IBM's Smarter Planet thinking," says Halim.

Rethinking computing from the ground up
This new computing paradigm turns data analytics on its head. In traditional data analysis, a static problem is loaded into storage and queries are run against it. In stream computing, data is pulled in on the fly from different sources. Each bit of data is handled by analytic software on servers optimized to digest and analyze it, so the computation can be much faster.

For example, an audio feed is sent to a translator which produces a stream of text. The text is then sent to the aggregator which filters it by topic: 30% sports, 70% financial news. Then it moves to the next processor for analysis and the next—all while new data streams in. Information that is relevant or "different" is picked up from the flood; irrelevant bits are sluiced away as "noise."

A platform that morphs around the data
While many databases can digest large amounts of information, InfoSphere Streams can perform analyses on an enormous, chaotic mix of data and deliver an ongoing stream of insights in microseconds. It continually adapts to the state and utilization of its computing assets and to the needs expressed by its users.

Depending on the type of workload, Research prototypes of InfoSphere Streams have been run on a variety of hardware architectures from generic servers to special-purpose architectures such as IBM Blue Gene. Continued on page 2

Page 1 | 2