Abstract

A Dagstuhl seminar on Compute-First Networking (CFN) was held online from June 14th to June 16th 2021. We discussed the opportunities and research challenges for a new approach to in-network computing, which aims to overcome limitations of traditional edge/in-network computing systems.
The seminar discussed relevant use cases such as privacy-preserving edge video processing, connected and automated driving, and distributed health applications leveraging federated machine learning. A discussion of research challenges included an assessment of recent and expected future developments in networking and computing platforms and the consequences for in-network computing as well as an analysis of hard problems in current edge computing architectures.
We exchanged ideas on a variety of research topics and about the results of corresponding activities in the larger fields of distributed computing and network data plane programmability. We also discussed a set of suggested PhD topics and promising future research directions in the CFN space such as split learning that is supported by in-network computing.

@Article{crowcroft_et_al:DagRep.11.5.54,
  author =	{Crowcroft, Jon and Eardley, Philip and Kutscher, Dirk and Schooler, Eve M.},
  title =	{{Compute-First Networking (Dagstuhl Seminar 21243)}},
  pages =	{54--75},
  journal =	{Dagstuhl Reports},
  ISSN =	{2192-5283},
  year =	{2021},
  volume =	{11},
  number =	{5},
  editor =	{Crowcroft, Jon and Eardley, Philip and Kutscher, Dirk and Schooler, Eve M.},
  publisher =	{Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik},
  address =	{Dagstuhl, Germany},
  URL =		{https://drops.dagstuhl.de/opus/volltexte/2021/15570},
  URN =		{urn:nbn:de:0030-drops-155706},
  doi =		{10.4230/DagRep.11.5.54},
  annote =	{Keywords: Distributed Machine Learning, distributed systems, edge-computing, in-network computing, networking}
}