What We Talk About When We Talk About Networks

Networks, and the applications they support, sometimes treat each other as strangers. By shaking things up a bit—expressing networked systems as compositions of small, pure functions and making their dataflow a first-class consideration—we can often achieve friendlier couplings across the stack, to the benefit of performance, robustness, and understandability. This approach has proved helpful in several contexts: networking algorithms learned "in situ," feeding data from deployment back into training; real-time video conferencing, especially for musicians and actors during the pandemic; image compression in a distributed network filesystem; and a serverless computing framework that lets software burst to 10,000 cores that we have used for software compilation and testing, film-scale 3D rendering, video encoding and inference, and other jobs. In ongoing work, we're building a "functional" operating system that enforces a separation between IO (declared to the OS) and computation (reproducible by default). This suggests an end-to-end argument for serverless computing, shifting the service model from “renting CPUs by the second” to “providing the unambiguously correct result of a computation.” Holding infrastructure accountable to these higher-level abstractions could permit agility and innovation on other axes.


Speaker

Keith Winstein

Associate Professor of Computer Science and, by courtesy, of Electrical Engineering @Stanford

Keith Winstein is an associate professor of computer science and, by courtesy, of electrical engineering at Stanford University. His research group creates new kinds of networked systems by rethinking abstractions around communication, compression, and computing. Some of his group’s research has found broader use, including the Mosh (mobile shell) tool, the Puffer video-streaming site, the Lepton compression tool, the Mahimahi network emulators, and the gg lambda-computing framework. Winstein previously served as a staff reporter at The Wall Street Journal and worked at Ksplice, a startup company (now part of Oracle) where he was the vice president of product management and business development and also cleaned the bathroom. (https://cs.stanford.edu/~keithw)

Read more

Date

Wednesday Mar 29 / 02:55PM BST ( 50 minutes )

Location

Whittle (3rd Fl.)

Topics

performance network infrastructure

Share

From the same track

Session performance

Performance: Adventures in Thread-per-Core Async with Redpanda and Seastar

Wednesday Mar 29 / 11:50AM BST

Thread-per-core programming models are well known in software domains where latency is important.

Speaker image - John Spray

John Spray

Storage Engineering Lead @neon.tech, Formerly Redpanda, Inktank (Ceph), Whamcloud (Lustre)

Session performance

Providing a Personalized Experience to Millions of Users @BBC

Wednesday Mar 29 / 01:40PM BST

Personalization can be a very effective and impactful way for your organisation to attract new customers as well as to retain your existing customer base – if you get it right!

Speaker image - Manisha Lopes

Manisha Lopes

Principal Software Engineer @BBC

Session performance

Adventures in Performance

Wednesday Mar 29 / 10:35AM BST

The end of the economical version of Moore's Law (exponential decrease in transistor unit costs), the advent of metered cloud computing, and the shift of the economy toward digital goods and SaaS-based business models has shifted performance engineering from the fringes ("CPU progress will fix it

Speaker image - Thomas Dullien

Thomas Dullien

Distinguished Engineer, Mathematician, World-Renowned Security Researcher

Session performance

WebGPU is Not Just About the Web

Wednesday Mar 29 / 04:10PM BST

Since OpenGL has been deprecated, writing portable native apps that leverage the massively parallel computing capabilities of GPUs has become tedious because no single graphics API can fit all targets.

Speaker image - Élie Michel

Élie Michel

Research Scientist @Adobe