Group communications appear in many recent datacenter applications. These applications, however, do not benefit from multicast due to the absence of efficient systems that support large-scale multicast sessions, minimize state at switches and reduce bandwidth overheads. We propose a new architecture, called Orca, that addresses the challenges of multicast in datacenter networks. The proposed approach carefully divides the state and tasks of the data plane among switches and servers in order to realize efficient multicast services in datacenters, by partially offloading the management of multicast sessions to servers. We implemented Orca in a testbed to show its practicality in terms of throughput, consumption of server resources, packet latency and the impact of server failures. We also implemented a sample multicast application in our testbed, and showed that Orca can reduce its communication time by up to 7.33X, through optimizing the data transfer between nodes using multicast instead of unicast. In addition, we simulated a datacenter consisting of 27,648 hosts and handling 1M multicast sessions, and we compared Orca versus the state-of-art system. Our results show that Orca reduces the switch state by up to two orders of magnitude, and the communication overhead by up to 19X, compared to the state-of-art.
- ConferenceOrca: Server-assisted Multicast for Datacenter NetworksIn Proc. of USENIX NSDI 2022