Information Management and Semantic Search
Information and knowledge are critical to Services Science. In the process of creating, marketing, and delivering service a service provider generates large volume of unstructured information in the form of documents, presentations, spreadsheets, email, and so on. This content presents a rich opportunity and a challenge for information management, mining, and semantic search. We are investigating algorithms for ontology development, semantics search, and automated extraction of concept-based overviews. We are applying these algorithms to large scale real life content generated in the process of services marketing and delivery.
Large Scale Systems and Infrastructure Management
Today's large scale systems are quite complex and yet as a part of the business-value driven optimization there is a need to transform them continually to more and more efficient states. It is not unusual for an enterprise to operate tens of 1000's of computing systems, running very large instances of middleware and applications, using hundreds of tera-bytes of data residing in very large storage systems and file servers. Along with the hardware and software complexity, there are also recoverability issues, change management processes, help-desk services, compliance procedures, outsourced operations, and so on. Transformation of such a complex system is a significant challenge whether the transformation is a part of an outsourcing service deal or ongoing enhancements planned by the in-house management. We are developing rule-based technologies to use "how-to" knowledge and best practices to automate design of IT transformation. We are exploring knowledge acquisition and representation methods, design verification algorithms, and machine learning techniques as applied to large scale systems. The transformation also gives raise to the need to verify that the transformation complies with service agreements.
Policy-Based Storage and Data Management
We conduct leading edge research in policy-based management of storage and data. We have developed policy-based management language and its implementation to manage billions of files in enterprise-scale file systems. A significant challenge in very large file systems is the efficiency of the search mechanism to determine the candidate files that are subject to policy actions - scanning through billions of files is not a viable option. Our algorithms predict policy actions ahead of time and significantly minimize the overhead. While continuing to pursue efficient and effective policy-based management implementations, we are also addressing the long standing need to associate data & storage with applications and vice versa. The classic UNIX-style I/O interface leaves a semantic gap between the user and the data used. We are developing algorithms and system design to make the semantic connection and thus establish provenance of data in the framework of business applications.