HyFlow


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.

 

Research Areas


  • Theory of Transactional Memory

  • Scalable Transactional Memory Implementations

  • Hardware/Hybrid Transactional Memory

  • Transactional/Composable Data Structures

  • Fault-Tolerant Transactions for Multicore and Distributed Systems

  • Geo-scale Transactions

TM Extentions for Low Level Semantics

Performance of the NOrec STM algorithm's two versions (NOrec and S-NOrec) on the Vacation application of the STAMP benchmark suite. The algorithms are integrated into our version of the GCC compiler, which instruments transactions using an innovative technique that captures application's low-level semantics to reduce abort rate. For details, see our SPAA 2016 paper.

 
M2Paxos - Making Fast Consensus Generally Faster

Maximum attainable throughput of M2Paxos against competitors including EPaxos, Generalized Paxos, and Multi-Paxos under varying number of nodes. Application workload includes write commands issued on keys, with 100% command locality. For details, see our DSN 2016 paper.

Recent Selected Publications


All papers

 
This work is supported in part by US National Science Foundation under grants CNS 0915895, CNS 1116190, CNS 1130180, and CNS 1217385, and AFOSR under grants FA9550-14-1-0163 and FA9550-14-1-0143. Any opinions, findings, and conclusions or recommendations expressed in this site are those of the author(s) and do not necessarily reflect the views of the National Science Foundation or AFOSR.