Lambda Days 2018 - Daniel Krzywicki - Scaling functional multi-agent computations (...)

Thanks! Share it with your friends!

You disliked this video. Thanks for the feedback!

Added by admin
11 Views
Register to Lambda Days newsletter to receive updates: http://www.lambdadays.org/lambdadays2018/#contact

Daniel Krzywicki
I am a Tech Lead at FaberNovel Technologies, where I work on hypermedia REST APIs, semantic web tooling, and distributed systems, mostly in Scala. I'm also a Ph.D. student at the Department of Computer Science of AGH-UST. My research interests include agent-based computations, functional programming and distributed computing.

Scaling functional multi-agent computations with reactive streams
Multi-agent systems are often used to simulate real-word phenomena or in optimisation algorithms. Traditionally, relevant software has often relied on coarse grained parallelism, with little or no concurrency.
However, large numbers of agents in such use cases mean that more fine grained parallelism is needed to fully scale on modern manycore hardware. Moreover, higher concurrency has a value at the domain level, as it allow to more closely simulate our (almost) "continuous" world.
In this talk, we first show how to use functional patterns to decompose a multi-agent simulation in a way which allows it to be run on execution models of varying concurrency and parallelism. Then, we see how such execution models can be generalized into a streaming one. Finally, we describe an Akka Streams implementation which allows to achieve fine-grained concurrency and parallelism for greater performance and efficiency.
Category
Lambda Day 2018

Post your comment