Fantastic Data designed and built the distributed data cache for the DARPA Sensor Information Technology (SensIT) program. This innovative approach provided data reliability without permanent storage by efficient maintenance of redundant data caches. The distributed data cache optimizes power consumption, provides data survivability across the network, and conserves bandwidth. The distributed data cache anticipates data needs and efficiently distributes data using an innovative multicast protocol to redundant caches on the nodes that are most likely to require the data.

The redundant caches eliminate the need for permanent data storage, thus eliminating one of the most operationally costly components of the nodes, and allows the development and deployment of significantly cheaper, hardier, and less power hungry devices. At the same time, the distributed data cache provides full object-relational database flexibility (transactions, queries, extended data types) to allow the distributed processing of application specific data into useful information for military operations.

The location and content of caches are determined automatically in response to data access requirements (latency, reliability, and survivability) and operating conditions (power, bandwidth, and node density). The data needs of the individual nodes of the network are determined by collecting and analyzing the transactions and queries performed against the database by application programs to form a “data interest” or “query” for the node. As data interests are dynamically altered, data stored on the nodes and data flowing between the nodes is adjusted to correctly reflect the new rules. Decisions are made in a distributed manner with emphasis on the ability to operate in a poor communications environment.

The collaborative tracker was developed to test the distributed data cache on a 25-node sensor network in Fall 2002. On each node, the tracking application expresses interest in detection data in an area about the node. This triggers the dissemination of detection data to the appropriate nodes where it is combined into tracks. Track data is then distributed using different rules to those nodes that have expressed interest in tracks. In this case, the distributed tracking application also requires track data, because it must reconcile its results with those of other nodes. Real time detection and tracking is a rigorous test of the distributed data cache, as the production of correct tracks is dependent upon fast and accurate data dissemination. Example tracks recorded during the field experiment are shown in the animations.

References
April 2, 2000 presentation

October 10, 2000 presentation

April 19, 2001 presentation

January 15, 2002 presentation

August 28, 2002 presentation

November 7, 2002 presentation