Track 1 Session 8

2:20 to 3:20 p.m. Thursday April 8, 2010

Cost-Effective Software Reliability Through Autonomic Tuning of System Resources

Often the systems our societies depend upon are built in such a way as to result in being too inflexible and intolerant to changes. The deployment of such systems in environments where change is the rule rather than the exception leads to situations where reliability and performance are strongly and negatively affected. In particular, upper bounds and worst-case analyses do not match those environments. Truly effective software reliability approaches require a precise characterization of the allocation of resources over time. In other words, unwanted emergent behaviors can only be avoided if the systems are built with a finer-grained control of the redundancy degree. In this presentation, we discuss this problem, summarize the lessons we learned by investigating it and provide an example of a compliant system – adaptively redundant data structures. In particular, we show how such systems allow software reliability to be guaranteed cost-effectively through autonomic tuning of the employed redundancy.

Key Words: Software Reliability, Software Engineering, Resilient Computing, Adaptability, Redundant Data Structures, Autonomic Computing

Vincenzo De Florio

University of Antwerp and Institute for BroadBand Technology

Belgium