Optimized SRv6 Multicasting for Network Assisted Publish-Subscribe Systems

The Internet is poised to support new classes of industrial applications, which are driven by large numbers of sensors and actuators generating and consuming sensor data streams. A publish-subscribe (pub-sub) paradigm is better suited to handle such large scale data dissemination than a client-server approach. SRv6 provides a convenient mechanism to build and maintain network-efficient, shareable data distribution trees for pub-sub brokers. We study optimal algorithms for setting up different types of multicasting in SRv6 systems. In particular, we show both theoretically and experimentally, that splitting the multicast streams into multiple sub-streams, as well as using end-to-end (application layer) coding without any network participation provides significant benefits compared to traditional single stream multicasting.