Live-traffic gives information-centric networking a boost
ICN addresses content by name rather than by the IP address of the server hosting it. The technology has been in development for about a decade, re-emerging earlier in 2018.
The trial provided early field data about how ICN performs in a real-world-like environment. An EU project established in 2014, POINT, has worked towards demonstrating IP-over-ICN on a test bed covering IPTV, IP telephony, and vanilla Internet access.
Code for POINT was first released in 2016 to operators and researchers.
In POINT closed trials, Cypriot operator PrimeTel tested the technology with real users accessing video streams either as IPTV or as HLS, HTTP Live Streaming.
As this publication at arXiv, released last Friday April 20th, explained, POINT is designed to let network operators deploy ICN with a minimum of disruption: customer equipment, and the routers/gateways connecting operators, remain unchanged.
By ICN-enabling the network, the video content follows a publish-subscribe model in which a show is accessed by name rather than (for example) URI. The paper explained that there are three core functions supporting this: Rendezvous matches publisher and subscriber nodes; Topology Management calculates and encodes paths between nodes (stored as Forwarding Identifiers, FIDs); and the Forwarding Node (FN) function that passes data based on FIDs.
“FIDs are included in packet headers, allowing FNs to forward packets with a few bitwise operations, without requiring routing tables or any routing state. Consequently, the PURSUIT ICN architecture can enable stateless multicast switching and native anycast.”
In the 2017 closed trial (staged between November 20th and December 1st, 2017), customers visited PrimeTel’s offices to compare video delivered over the operator’s conventional IP infrastructure to their experience using the ICN-capable POINT infrastructure, with the test operators adding congestion to the core network.
The test gathered network data, users’ self-reported experience, and recorded user sessions to look for events like video artefacts or user responses to events (such as trying to pause and resume, if there was an interruption).
For HTTP live-streaming, the “burn test” applied to the services involved taking down the main server, then forcing a lower-bitrate surrogate server to deliver the stream.
Over the IP network, the paper explained, the failure afflicted users with “a long duration” of pixellation, but the POINT ICN network the stream switched servers “without any noticeable effect to the end-users”.
The IPTV test used two links between switches serving customers. Here, the paper said, the ICN-based network had an automatic advantage: where conventional spanning tree networks demands only one link is live at a time, in POINT both links are active at once.
“In the IP case, we brought down the primary interface, which led to recalculation of the spanning tree and re-establishment of the IGMP snooping state, causing major viewing disruption”, the paper explained – and the same steps took place when the link was brought back up.
On the ICN infrastructure, there were “no noticeable disruptions” when links were dropped or brought back.
The paper noted that the groups involved in POINT plan further analysis of data from the closed test before a two-week open test on PrimeTel’s live network.
The analysis was conducted by PrimeTel, InterDigital, CTVC UK, Intracom SA Telecom Solutions, Germany’s RWTH Aachen University, and the Athens University of Economics and Business. ®