Lock-free Concurrent Data Structures and How to Model their Performance
Philippas Tsigas · 2019
Concurrent data structures provide the means to multi-threaded applications to share data. Typical designs of concurrent data structures are based on locks in order to avoid inconsistency due to concurrent modifications. Locks though introduce a sequential component in Amdahl's law. Lock-free algorithmic designs of concurrent data structures were introduced in the quest for better performance and scalability and are widely used in practice. Lock-free designs typically employ optimistic conflict control making performance analysis challenging. In this talk, I will describe recent efforts in analyzing their performance.