Authors: Róbert Attila Potys, Noman Mumtaz Ali (KTH), Ian Marsh and Flutra Osmani (SICS Swedish ICT)
Venue: IEEE/ACM 23rd International Symposium on Quality of Service (IWQoS), Portland, June 2015
DOI: 10.1109/IWQoS.2015.7404745
Year: 2015
Topic: IP Networking
Paper (IEEE Xplore) · Internet caching project · Robert Potys's thesis · Noman Mumtaz Ali's thesis
Summary
TCP was built for conversations between two hosts. Much of today's Internet is not a conversation but content retrieval: fetch this video, this update, this page, from wherever it is closest. Information-centric networking (ICN) redesigns the network around named data instead of host addresses. If data has a name, any node holding a copy, a cache, a peer or the origin, can serve it.
That needs a different transport protocol. NetInf TP is a transport for the NetInf ICN architecture. It is receiver-driven: the receiver asks for named chunks of an object and controls the pace. Because requests name data, not hosts, NetInf TP allows several senders and several receivers of the same data, and works naturally with caches along the path. Large objects in particular can download faster by pulling chunks from several sources.

Design
- Receiver-driven requests. The receiver keeps a window of outstanding chunk requests, like TCP's congestion window but on the receiving side.
- Reliability. The receiver detects missing chunks and re-requests them. The paper introduces a new type of retransmission mechanism suited to multiple sources, where a lost chunk may come again from a different node.
- Congestion control. The request window grows and shrinks with network conditions, so NetInf TP backs off under congestion like TCP does.
- Caching built in. Intermediate nodes can store chunks as they pass and answer later requests, shortening download times for popular content.
Results
The prototype was evaluated in simulation in a constrained environment, with several NetInf TP flows sharing a bottleneck with TCP flows. NetInf TP gave reliable transfer, reacted to congestion and co-existed with TCP, which the authors judged makes it a promising candidate for ICN transport.
Discussion
Receiver-driven transport moves control to the side that knows what it needs. That fits content retrieval and multi-source download, but raises its own problems: the receiver must estimate the path to several sources whose delays differ, and a cache that answers some requests changes the round-trip time mid-transfer, which can confuse congestion control.
Much of ICN's thinking went mainstream in other forms. CDNs and HTTP caching deliver named content from nearby caches. QUIC is not information-centric, but moved transport into user space, where new ideas are easier to deploy. Named Data Networking (NDN) continues as a research architecture with its own receiver-driven congestion-control work.
Future work
- Test on real networks and the NetInf prototypes, beyond simulation.
- Multipath, multi-source scheduling: choose which source serves which chunk based on measured delay.
- Fairness between ICN and TCP at scale, and with active queue management.
- Use for edge caching in mobile networks, where the nearest copy changes as users move.
The EU perspective
ICN was a major European research theme. NetInf was developed in the EU FP7 projects 4WARD (2008–2010) and SAIL (Scalable and Adaptive Internet Solutions, 2010–2013), led by Ericsson with SICS among the partners. See Considerations for a Network of Information. The EU–Japan GreenICN project and the IRTF ICN Research Group carried the work on. Today, the ideas live on in the EU's interest in content-aware edge networks under the SNS Joint Undertaking for 6G, and in net neutrality rules (Open Internet Regulation (EU) 2015/2120), which allow caching for efficiency but not to discriminate between content providers.
References
- R. A. Potys, N. M. Ali, I. Marsh and F. Osmani, "NetInf TP: a receiver-driven protocol for ICN data transport," IEEE/ACM IWQoS, 2015.
- B. Ahlgren et al., "A survey of information-centric networking," IEEE Communications Magazine 50(7), 2012.
- V. Jacobson et al., "Networking named content," ACM CoNEXT, 2009.
- G. Carofiglio, M. Gallo and L. Muscariello, "ICP: design and evaluation of an interest control protocol for content-centric networking," IEEE INFOCOM Workshops, 2012.
- D. Kutscher et al., "Information-centric networking (ICN) research challenges," RFC 7927, 2016.
Citation
Róbert Attila Potys, Noman Mumtaz Ali, Ian Marsh and Flutra Osmani, NetInf TP: A receiver-driven protocol for ICN data transport, IEEE/ACM IWQoS, 2015, Paper.
BibTeX
@inproceedings{potys2015netinftp,
author = {Potys, R{\'o}bert Attila and Ali, Noman Mumtaz and Marsh, Ian and Osmani, Flutra},
title = {{NetInf TP}: A receiver-driven protocol for {ICN} data transport},
booktitle = {IEEE/ACM 23rd International Symposium on Quality of Service (IWQoS)},
year = {2015},
doi = {10.1109/IWQoS.2015.7404745}
}
Related student theses
- Robert Potys: Implementation and evaluation of NetInf TP, an information-centric transport protocol
- Noman Mumtaz Ali: Implementation and evaluation of NetInf TP, an information-centric transport protocol
More in IP Networking: ← Understanding cache dynamics in the network | Hierarchical caching simulation with visualisation →