The Hyflow project is developing concurrency control abstractions, protocols, and mechanisms for current and emerging multicore architectures, cluster systems, and geographically distributed systems. A particular focus is to understand what concurrency control abstractions promote high programmability for these architectures, and how to support those programming abstractions with high performance, scalability, and dependability. A closely related focus is to build open-source experimental systems that embody the techniques, as well as incorporate them into existing open-source infrastructures.
Ongoing efforts are exploring high performance fault-tolerant transactional memory for cluster systems, high performance software transactional memory for multicore architectures, scalable transactional data structures, and scalable transactional protocols for geographically distributed systems.
Recent Selected Publications
- "Opacity vs TMS2: Expectations and Reality", S. Hans, A. Hassan, R. Palmieri, S. Peluso, and B. Ravindran, 30th International Symposium on Distributed Computing (DISC), September 2016, Paris, France
- "Disjoint-Access Parallelism: Impossibility, Possibility, and Cost of Transactional Memory Implementations", S. Peluso, R. Palmieri, P. Romano, B. Ravindran and F. Quaglia, The 34th Annual ACM Symposium on Principles of Distributed Computing (PODC), July 2015, Donostia-San Sebastian, Spain