Monitoring at Any Cost
Abstract
Rules are integral to our social and business reality. In many domains, rules are sufficiently precise and straightforward that compliance can be monitored automatically. Abstractly, a monitor takes a rule as an input, expressed in some specification language, and then processes a stream of timestamped events while detecting and outputting violations of the rule as early as possible. In practice, monitors face challenging real-world requirements like handling immense volume of streamed data that arrives at a high velocity. State-of-the-art monitors struggle to meet this requirement.We shall study and develop monitors that scale in real-world settings by learning and exploitinginformation about the event stream and by explicitly trading off output details and precision in order to optimize the monitors’ operation. We outline three monitoring approaches, ordered with respect to their ability to handle increasing volume and velocity of streams of events. We will develop parallel monitors that efficiently utilize available resources, as well as monitors that can improve their efficiency by omitting details from their output when necessary. Finally, if all else fails, we will allow the monitors to be incorrect with a very low probability, while still providing useful information in real time.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Sep 11, 2017
- Source ID
- FA95501710306
Entities
People
- David Basin
Organizations
- Air Force Office of Scientific Research
- ETH Zurich
- United States Air Force