Programming Technologies and Software Engineering

This broad research area covers programming models and languages, software processes, software maintenance, software testing and verification for a spectrum of applications ranging from business applications to scientific computing, and sequential to distributed and high performance computing.

Software Maintenance:
Maintenance of software constitutes a significant fraction of IT spending in typical enterprises. The goal is provide methodologies, tools, metrics for software quality and productivity that would provide a systematic basis for software maintenance.

Highlights of some of the projects in this direction are given below.

1. Model extraction from legacy applications
2. Tool-assisted software porting

Programming Languages and Systems: Business and scientific computing businesses are increasingly becoming collaboration-oriented and event-driven, both internally and externally. They have started deriving business benefits in terms of operational agility and efficiency, superior supply chain management enabled by streamlined interaction with partners, and suppliers, faster time to market, and addressing new market opportunities and competitive threats.

There is also significant interest and momentum around standardization efforts across industries. On the technological side, the device scaling predicted by Moore's law is no longer a viable option for increasing the clock frequency of future processors at the rate that had been sustained for last two decades. Hence, future systems are rapidly moving from uniprocessor to multi-processor configurations, so as to exploit parallelism for increased performance.

Current object oriented languages that offer concurrent programming, have revealed several drawbacks of parallel programming at the level of unstructured threads with lock-based synchronization, making these ill prepared for the challenge. Our focus is on devising new programming tools and systems for open standards-based business processing and scientific processing.

The following are our ongoing and recent projects in this area.

1. Models for complex event driven systems
2. Percolator: analyzing Java APIs for correct usage
3. X10: A programming language for high-performance computing
4. Symphony: Tool for decentralizing BPEL applications

Model driven architecture for end-to-end business applications
We are primarily conducting research in the use of formal models to explicitly define the structure and behavior of a business system, to monitor, analyze, and improve performance and leveraging models "as a recipe for construction" of business's IT systems. In this endeavor of closing the business to IT gap, we are researching several practical challenges associated with Model Driven Architecture (MDA). Some of the key ones are model based composite application building and execution, meta-models for incorporation of business rules and policies, meta-models for human-interactions, and change management issues.

Distributed Software Development
Business reasons are increasingly causing software projects to be distributed across the globe. However, a multi-site project brings with it several challenges like inadequate communication between teams, lack of information about remote sites, differences in processes etc., which disrupt the intrinsically collaborative nature of software development. The problem is compounded by the fact that most software tools and techniques in use today largely ignore the needs of distributed teams. We have been looking at global software development practices to understand problem areas and propose solutions that could be of help.

We have focused particularly on how to effectively manage requirements across sites. The elicitation of business requirements, refinement of high-level requirements to more concrete forms, propagation and management of changes in requirements - all of these require rich communication and effective co-ordination between the customers, analysts, system engineers and developers. Remoteness and time zone differences strain these scenarios, leading to excessive re-work, delays and cost escalations.

The following are our ongoing and recent projects in this area.

1. EGRET: Eclipse-Based Global Requirements Management Tool

Software Verification
Formal methods have a great potential as powerful means for the specification, early debugging of requirements, testing, validation and certification of software. They have been successfully applied in a spectrum of industrial applications and are even becoming an integral part safety standards. We are exploring the use of IBM’s industrial strength tools such as RuleBase for verification of software. We have been building methodologies for verifiable SystemC and thus build suite for verifiable system-on-chip designs.