OpenDaylight Summit 2016

Typically ODL applications that render to southbound OpenFlow capable switches assume they have complete ownership of all OpenFlow tables on the switch being programmed. The assumption is that if it's necessary for another application hosted by the same ODL controller to program flows, then that application do so on another switch, and packets will be passed back-and-forth between the switches. An example of this is the GBP+SFC integration proof of concept implemented in the ODL Lithium release. GBP controlled the endpoint groups in its own set of switches, while SFC controlled the service chaining in its own, separate set of switches. This worked ok until we wanted to integrate SFC and GBP into OPNFV. The OpenFlow switches are now controlled by OpenStack, and there will be 1 switch per compute node, meaning multiple ODL applications now need to share the same OpenFlow switch.