Session Date/Time: 09 Nov 2023 08:30

coinrg

Summary

This meeting of the coinrg research group covered a variety of topics related to computing in the network, including end-to-end principle considerations, automating in-network computing deployments, machine learning at IoT gateways, use case analysis, directions for computing in the network, transport protocols for in-network applications, and a new project focused on AI-driven personalized media. The discussions highlighted the ongoing evolution of the field and the need for further research and collaboration.

Key Discussion Points

End-to-End Principle: Ike presented a distillation of discussions around network computing and the end-to-end principle, highlighting different mental models, functionality considerations, and interpretations of the principle. The presentation sparked a discussion about how to move forward and whether the constraints of the existing internet architecture should be a starting point.
Automated In-Network Computing: Heyouong presented ClickInc, a system for automating the development and deployment of in-network computing applications in data centers, focusing on decoupling network operation from application development.
Machine Learning at IoT Gateways: Ming Wang Zhang presented work on applying in-network classification for traffic analysis at smart IoT gateways, focusing on resource constraints and continuous defense.
Use Case Analysis: Yong presented an update on the use case analysis draft, now refocused on research challenges. The presentation prompted feedback about the breadth of the scope and the need for more concrete examples.
Directions for Computing in the Network: Doug outlined the draft, discussing different types of network computing systems, their characteristics, and associated challenges. The presentation garnered positive feedback on the angle and concept clarification but also suggestions for a clearer problem statement.
Unified Transport Protocol: Heyouong presented a new draft on the requirements for a unified transport protocol for in-network applications, focusing on remote procedure call-like patterns and the need to address the new end-to-mid-point-to-end model.
Evolving Cooperating Layered Architecture for SDNs: Carlos presented an update on RFC with an evolution of data awareness, and compute layers in SDNs, adding a compute stratum and data analysis plane.
AI for Me: Daniel introduced the AI for Me project, a collaboration with the BBC focused on personalized media distribution, highlighting the significant computational challenges involved and areas of exploration.

Decisions and Action Items

Use Case Analysis Draft: Yong will narrow the scope of the draft and provide more specific examples in the next iteration, as well as incorporate comments on structure and organization.
Directions for Computing in the Network Draft: Doug will consider adding a clearer problem statement and incorporate feedback.
Last Call for Use-Case Analysis Document: Finish date set for 2 weeks after IETF week.

Next Steps

Continue discussion on the mailing list.
Explore collaboration opportunities between different efforts within the research group.
Consider further investigation and concrete examples relating to collective communications and transport abstractions.
Daniel from the AI for Me project indicated they will consider submitting an Internet Draft or a more focused discussion on a specific area of compute in the network.