Berkeley C++ Meetup #7: CAF - Transparent Scaling from IoT to Data Center Apps

C++ Actor Framework - Transparent Scaling from IoT to Data Center Apps

Presented by Matthias Vallentin

Wednesday, November 2nd, 2016
7:00 p.m.
ICSI Lecture Hall

Abstract:

Today's compute landscape consists of heterogeneous systems:large-scale clusters of commodity machines at one end of the spectrum, special-purpose hardware like GPUs at the other end. The diversity of modern many-core architectures keeps increasing and brings us unprecedented performance at the cost of standardized memory access latencies. Meanwhile, low-powered IoT devices creep into our daily lives without an unforeseeable end. This talk introduces C++ Actor Framework (CAF), a uniform abstraction to write reliable and efficient programs across this entire spectrum. At its core, CAF provides a message passing abstraction in terms of the actor model, which offers a paradigm to unify concurrency and distribution.In the actor model, computational entities - called actors - execute independently and in parallel. Using unique, location-independent addresses,actors communicate asynchronously solely via message passing.Because actors do not share state, data races cannot occur by design. After introducing basic concepts and the API, I will present an evaluation that demonstrates that CAF offers superior performance compared to other actor runtimes (e.g., Akka/Scala, Erlang, Charm++, etc.) while providing a type-safe high-level abstraction.

Author:

Matthias Vallentin is a post-doc at UC Berkeley, where he works on distributed systems to improve large-scale network forensics. Before that, he was a research scientist at the International Computer Science Institute (ICSI),where he continues to collaborate with the security group. Prior to that, he finished his Ph.D at UC Berkeley, where he gained extensive exposure to modern C++ and scalable distributed systems.

RSVP & Details: https://www.meetup.com/ACCU-Bay-Area/events/233766624/