Abstract

In this talk, I will first present “asynchronized concurrency (ASCY),” a paradigm consisting of four complementary programming patterns. ASCY calls for the design of concurrent search data structures to resemble that of their sequential counterparts. I will show how ASCY leads to implementations which are portably scalable: they scale across different types of hardware platforms, including single and multi-socket ones, for various classes of workloads, and according to different performance metrics, including throughput, latency, and energy.

I will then introduce our ongoing work on OPTIK , a concurrency framework that facilitates the design and implementation of ASCY -compliant data structures. OPTIK relies on the well-explored ideas of versioning and validation: a thread acquires a version number, optimistically accesses the data structure, and then locks and validates the version to detect concurrency. I will show how OPTIK can be used to easily make lock-based data structures more optimistic and more scalable.

(Asynchronized concurrency appeared in ASPLOS ‘15)