With billions of consumer Internet devices, medical instruments, and scientific instruments spawning petabytes of data, there are exciting opportunities to create intelligent behavior and new scientific insights. Whilst data has always been extremely valuable, with the explosion in sensing and data acquisition of all types, it now stands poised to eclipse the importance of computation in computer science and informatics. Unfortunately, Petabyte systems remain the tools of elite large corporations with prohibitive cost and complexity. We are interested in new ideas which simplify use and create new capabilities for Exabyte systems of the future – and will enable a personal Exabyte computing revolution. Research areas include systems software (scale, resilience, energy-efficient), novel programming (e.g. declarative), and the impact of emerging future storage technologies (e.g. PCM and molecular storage) on data center and system architecture. Current efforts include:- Blockus - Scalable, Efficient Data-intensive Computing based on blocks (efforts include building a performance model, a benchmark suite, new system architecture, intelligent scheduling and prefetching)
- Global View ResIience (GVR) - a flexible data-oriented approach for reliable computing in Exascale systems (efforts include distributed arrays, runtime, novel checkpointing to non-volatile memory, flexible error protection, compression, and co-design for recovery)
|
|
