Explore a novel architecture for efficient multicast in datacenter networks through this conference talk from NSDI '22. Dive into Orca, a system that addresses scalability challenges by dividing tasks between switches and servers, and partially offloading multicast session management to servers. Learn how this approach significantly reduces switch state, minimizes bandwidth overhead, and maintains constant processing overhead without limiting multicast session size. Examine the implementation results from both a testbed and large-scale simulation, comparing Orca's performance to state-of-the-art systems. Discover how Orca optimizes data transfer in multicast applications, potentially reducing communication time in datacenter environments.
Orca - Server-assisted Multicast for Datacenter Networks
Overview
Syllabus
Intro
Group Communication in Datacenters
Multicast in Datacenter Networks Can benefit from multicast
Challenges of Multicast Forwarding in DCNS Large-scale multi-tenant environmen
Our Goal
Our Approach
Orca Labels
Orca: Leaf Labels • Leaf switch forwards multicast packets to Orca Agent
Orca: Source Label
Orca Architecture
Fault-tolerant Forwarding
Evaluation using Simulations A Leaf-spine topology with 27K hosts
Orca vs Elmo
Evaluation in a Testbed
Benefits of Orca to Applications
Agent Scalability
Conclusions Orca is an efficient multicast system for DCNS • The key idea is to move parts of forwarding to servers
Taught by
USENIX
