Peter J. Clarke

Keywords of research areas:

Software Testing, Testing Autonomic Systems, Model-Based Testing, Model-Driven Software Development, Modeling User-Centric Communication.

Testing Autonomic Systems:

Problem: How to ensure that after a self-management change has been made to an autonomic computing system at runtime, the system still functions correct with respect to its original requirements?

Approach: We have proposed an implicit self-management characteristic referred to as self-testing. To perform self-testing at runtime we have identified two approaches: safe adaptation with validation and replication with validation. Both approaches use the autonomic systems structure as outlined in the IBM architectural blueprint for autonomic computing.

Results: We have developed a conceptual framework for self-testing autonomic systems. Our framework interacts with the autonomic components in a system to be testing. These components include: orchestrating test managers, touchpoint test managers, and test knowledge sources. To show the feasibility of our approach we have implemented two prototypes of autonomic applications with capabilities such as self-configuration, self-optimization and implicit self-testing.

Challenges: Apply our approach of self-testing to more complicated autonomic computing systems and perform empirical studies to identify the practicality of our approach in real systems.

Modeling User-Centric Communication Services:

Problem: Current user-centric communication-intensive systems are developed using a stovepipe approach that results in a lengthy and costly development cycle. How can user-centric communication services be modeled and realized in a more efficient manner with respect to development time and cost?

Approach: We propose the Communication Virtual Machine (CVM) technology that provides a model-driven platform for formulating, synthesizing, and executing new communication services. When a new (communication) service is needed, a model called communication schema that specifies the requirements and work flow for the service is built using a modeling tool and then realized by CVM. The CVM technology also includes the use of one or more mediators that interface with heterogeneous data sources. Data retrieved form these data sources can be dynamically packaged into a single entity known as a form. The form structure may be predefined or it can be dynamically created by the user of CVM.

Results: We have developed a CVM prototype form the healthcare domain that allows users to model and realized scenarios from the healthcare domain. We use the communication modeling language (CML) to model the communication services aspect of the scenarios.

Challenges: Expand CML to allow for the modeling of the workflow aspect of the communication-intensive applications by using constraints.