QoS management techniques. We investigate adaptive techniques, algorithms, and control for resource allocation, scheduling, and service-level assurance in dynamic, heterogeneous computing environments. Our research interests in this area include:
- Total cost of ownership, including cost of inference
- System and agent orchestration
- Resource management, scheduling, and control
See our recent QoS management publications for more details.
Predictive modeling and quantitative evaluation. We formulate robust predictive frameworks and cost models, based on AI/ML and analytical methods, to assess and learn the performance and reliability of distributed computing systems in the presence of faults, queueing, synchronization, and caching. Our research interests in this area include:
- Deep surrogate models
- Simulation
- Queueing theory/Markov chains
See our recent Predictive modeling and quantitative evaluation publications for more details.
Dependability and fault-tolerance. We develop resilient architectures and fault-tolerance mechanisms to predict and manage faults in AI/ML deployments and edge/fog computing systems, minimising any reduction in QoS arising from faults and selecting remediation actions. Our research interests in this area include:
- Anomaly detection
- Resilient architectures
- Fault remediation
See our recent Dependability and fault-tolerance publications for more details.