The Event Web: Sense and Respond to Critical Conditions
This is the first in a series of articles about the Event Web (EW), a thin layer on top of the World Wide Web that is continuously active, monitoring information sources and responding appropriately as conditions change. A new type of application called Sense and Respond (S&R) applications, that continuously respond to critical conditions, can be built on top of the EW.
The series will discuss the following technologies and solutions:
- Service-oriented architectures (SOA) and event-driven architectures (EDA)
- Relationships that the components of these architectures have to technologies such as data stream management systems, business intelligence systems, and rule-based systems
- Ways of implementing event-driven applications using Java and .NET
- A variety of S&R applications, ranging from intrusion detection to trading
The EW is a project in the Infospheres Group at Caltech. We hope this series encourages developers to form a community to build tools that make the EW a reality. This initial article presents the motivation for developing S&R applications but does not present implementation details; the articles further in the series will.
Sense and Respond Applications
S&R applications amplify an enterprise's capability to respond to threats and opportunities. The ability to respond appropriately to critical conditions is crucial to survival. In nature, for example, a zebra that doesn't run away from a lion dies. Conversely, a zebra that continuously runs away from non-threats can die from wasted energy. S&R applications have actuators or responders that execute actions when the applications detect critical conditions, just as animals have muscles that help them respond.
Photo courtesy of Corbis.com
Click here for a larger image.
Photo courtesy of Corbis.com
An enterprise S&R system consists of a network of distributed S&R systems, each serving a specific role. S&R applications share many of the behaviors of groups that respond collectively to their environments. You can think of a pride of lions as an S&R system consisting of multiple lions, each of which, in turn, is an S&R system.
Living things form models for the ways in which components of their environments behave: A zebra has an innate model of a lion's behavior; cattle that graze near a railroad track learn a model of a train's behavior. Similarly, S&R systems have models of their environments, which may be either specified by users or learned by the system. Machine learning algorithms help the systems learn critical conditions.
Sensing the Environment Beyond the Enterprise
An enterprise's threats and opportunities arise from events both inside and outside the organization. Competitors, governments, news organizations, and markets can generate opportunities and threats. That's why S&R systems monitor events both outside and inside the enterprise. By contrast, traditional enterprise information systems focus on information only within the organization where the time, place, form, and accuracy of events are controlled.
The most challenging aspect of monitoring the environment outside the enterprise is that the data is not within the enterprise's control: Competitors do not structure data in schemas that suit the enterprise, nor do they deliver data on schedules that the enterprise prescribes. As a consequence, errors are inevitable: data from the environment may be imprecise (as in natural-language text), incorrect, or delayed. A system may fail to detect a genuine threat or opportunity, and it may incorrectly perceive threats and opportunities that do not exist. S&R takes this reality into account and places greater emphasis on asynchrony and imprecision than traditional EAI does.