Context-aware Policy Matching in Event-driven Architecture
Shao-you Cheng
Wan-rong Jih
Jane Yung-jen Hsu
r93070@csie.ntu.edu.tw jih@agents.csie.ntu.edu.tw yjhsu@csie.ntu.edu.tw
Computer Science and Information Engineering, National Taiwan University, Taiwan
Motivation
Applications for supporting pervasive computing and agility are the current trend in software development. The service-oriented architecture (SOA) enables connection between consumers and service providers in a loosely-coupled way to improve flexibility and extensibility. This architecture is static, which utilizes predefined sequences of actions and fixed policies. Ongoing processes cannot adapt to dynamic changes in the environmental conditions or context.
Imagine the situation where a real estate broker shows her client a house for sale, matching the preference profile pro-vided by the potential buyer. While the buyer likes the gen-eral location of the house, he considers it unacceptable due to the unexpected traffic noise from a nearby street. Instead of proceeding with the original plan to show another house on the same street, an experienced broker should adjust the plan in light of this additional constraint. The broker con-nects to the multiple listing service with her mobile device and downloads a newly listed house within minutes of the current location that better satisfies the client’s requirements. Such a scenario can facilitate the introduction of events into SOA (He 2003). Real-time changes are modeled as events, which in turn trigger changes of states for the work-flow to meet the business needs. In this scenario, events can be ”shows client a house”, ”add client’s requirements”, etc.
Introduction
This research explores the role of context-aware policy matching in an event-driven architecture (EDA). In partic-ular, a context-aware rule engine is adopted to derive con-clusions based on the current contexts and business policies. Figure 1 shows the functional modules of the prototype de-signed to demonstrate the advantages of the proposed ap-proach. In the Underlying Architecture layer, standard SOA is combined with EDA. The Context-Aware Rule Engine layer consists of three distinct agents for collecting prefer-ence profiles, ambient or context information, and dynamic events. All information collected will be forwarded to the Rule Engine, augmented with the Rule Repository and Con-text Ontology. Results from the Rule Engine will be given to the Action Agent that performs the desired sequence of
Copyright c° 2005, American Association for Artificial
Intelli-gence (www.aaai.org). All rights reserved.
actions. The following sections contain detailed description of each module.
Action Agent
Rule Repository Rule Engine Context Ontology
Profile Agent Context Agent Event Agent
SOA EDA
Context-Aware Rule Engine
Underlying Architecture
Figure 1: Context-aware policy matching framework
Event-driven Architecture
The SOA use a simple and clear interface to bind all par-ticipating software components and provides service reuse. However, it lacks proactive ability. Event notification is the core of EDA, it can tightly bind services, events, and con-sumers in a dynamic environment. Events are more likely to complement, not replace, services in SOA. Since EDA has been touted as the “the next big thing” on the horizon of software development methodology (Schulte 2004). It is generally believed that event-driven services under SOA gain benefits from the features of both architectures (Hanson 2005).
Conventional event-driven design is also referred to as message-based systems, where a message here is a form of an event. Specifications of Notification family and WS-Eventing define the standard Web services approaches for event handling. When events occur, as shown in Figure 2, service producers publish the messages which will be deliv-ered, e.g. via publish-and-subscribe, to the event consumers.
Context-Aware Rule Engine
The proposed framework utilizes the Jess rule en-gine (Friedman-Hill 2005) to provide inference results. Jess processes the rules and facts using the Rete algorithm (Forgy 1982). User preferences are represented as policy rules that are matched in the reasoning process.
Contexts refer to the various dynamic aspects of the en-vironment, for example, location, time, and people (Dey
Providers
Consumers
User
Profiles Sink Sink Sink Sink
… Service Message Channel and Content Matching Publish Subscribe Content Matching Content Matching … … Send Receive Receive Publish Send Subscribe Service Service Service Service
Policies
Sink Sink
…
Events
Product announcement Top newsNew arrival Special offer Season discount Stock price
… …
Figure 2: Message-based event-driven architecture 2001). Static rules cannot react to the dynamically chang-ing contexts. As a result, context-aware rules are defined to use dynamic facts in the knowledge base, which may result in different conclusions depending on the current context. In our previous work, we successfully implemented context-aware rule-based reasoning on Java-enabled mobile devices, such as the iPAQ, to perform access control of sensitive in-formation (Jih, Cheng, & Hsu 2005).
A context agent monitors and perceives any environmen-tal changes. Figure 3 shows that context information may be captured by various multi-modal sensors, such as GPS or RFID. The context agent filters the contexts that have been detected, passing the selected contexts to the rule engine. Similarly, the event agent perceives and filters the relevant events of its surrounding area, and the profile agent keeps track of the user’s preferences.
Context-Aware Policy Matching
Context-Aware Rule Engine
Rules Policies Rule Engine Fact KB Context KB Context Agent Contexts location people
time target activity Contexts place person time target action
Inference Services Processes Events Plans … … Context Ontology
Figure 3: Context-aware rule engine
Now, let’s examine the scenario of the real estate broker again. The buyer’s rejection of the current house generates two events. First, the buyer is not satisfied with the choice, so he needs additional recommendations. Second, the buyer
preference profile is incomplete, and should be modified to reflect his preference for a quiet location. The former event triggers the broker’s mobile device to request for additional listing from the server, subject to successful matching be-tween the updated preference profile and updated new list-ing.
When contexts are encoded in the rule sets, inconsistency in the terminology is not only confusing, but also leads to incorrect reasoning outcomes. Given that most useful con-texts are tightly related to the problem domain (e.g. real estate), ontology can be used to bridge the semantic gap among different vocabularies. For example, while location is a common component of context defined in many appli-cations, sometimes it is referred to as place, space, or area, and so on. A context ontology helps generate a complete model of the different contexts (H.Wang et al. 2004) for a given domain. A specific context might directly derive from a more generic one, aggregate to a complex context, or up to an abstract context. Moreover, the context ontology will support hierarchical views of contexts (Gu et al. 2004) to improve the reasoning power of the rule engine.
Conclusion
This paper describes the design of a proposed framework for deploying a context-aware rule engine to the event-driven services platform in order to provide agile and real-time ser-vices. The design uses an agent architecture and a rule en-gine for flexibility and scalability in software development. A context ontology is utilized to resolve inconsistent vocab-ularies in knowledge sharing and rule merging.
References
Dey, A. K. 2001. Understanding and using context.
Per-sonal Ubiquitous Computing 5(1):4–7.
Forgy, C. 1982. Rete: A fast algorithm for the many pat-terns/many objects match problem. Artificial Intelligence 19(1):17–37.
Friedman-Hill, E. 2005. Jess, the rule engine for the java platform. Sandia National Laboratories.
Gu, T.; Wang, X. H.; Pung, H. K.; and Zhang, D. Q. 2004. An ontology-based context model in intelligent environ-ments. In Proceedings of Communication Networks and
Distributed Systems Modeling and Simulation Conference,
270–275.
Hanson, J. 2005. Event-driven services in soa. JavaWorld. He, H. 2003. What is service-oriented architecture? O’Reilly Web services.XML.com.
H.Wang, X.; Gu, T.; Zhang, D. Q.; and Pung, H. K. 2004. Ontology based context modeling and reasoning using owl. In Proceedings of Workshop on Context Modeling and
Rea-soning(CoMoRea’04), 18–22.
Jih, W. R.; Cheng, S. Y.; and Hsu, J. Y. J. 2005. Context-aware access control on pervasive healthcare. In EEE’05
Workshop: Mobility, Agents, and Mobile Services (MAM).
Schulte, R. 2004. Event-driven architecture: The next big thing. Application Integration and Web Services Summit. Gartner, Inc.
