SDN Multicast Membership Control for Scalable Data Center Overlays
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multicast solutions in data center overlay networks require intermediate nodes to maintain state for each multicast address, leading to unnecessary processing burden on hypervisors, especially when only a small portion of hypervisors need to process multicast data frames, and existing VPN multicast approaches do not scale well due to differences in VM migration and attachment dynamics.
Innovation Solution
A centralized DC-SDN controller manages membership information of multicast groups, sends gratuitous messages for address mapping, and uses query and report messages to update group memberships, enabling proper multicast data frame processing and delivery without requiring multicast support from overlay edge nodes or changes to existing switches/routers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing multicast solutions are used in overlay networks, then multicast delivery can be achieved, but intermediate nodes must maintain state for each multicast address causing processing burden
Solution Approach 1:
The patent introduces a dedicated multicast router as an intermediary node that assumes the responsibility of maintaining multicast state and processing multicast data frames. This mediator handles the complexity of multicast management centrally, while overlay edge nodes and hypervisors are relieved of this burden. The multicast router receives multicast data frames from overlay edge nodes, maintains multicast group state, and forwards frames to appropriate members, thus resolving the contradiction between reliable multicast delivery and reduced processing burden on hypervisors.
2Productivity
If multicast is implemented in overlay networks, then efficient group communication is achieved, but existing VPN multicast approaches do not scale well
Solution Approach 1:
The patent segments the multicast functionality into distinct components: overlay edge nodes that handle encapsulation/decapsulation, a dedicated multicast router that handles multicast routing and state management, and the overlay network that provides virtualization. This segmentation allows each component to perform its specific function efficiently. The multicast router maintains scalability by managing multicast state independently of VM migrations, while overlay edge nodes handle local VM attachments. This modular segmentation enables the system to scale better with dynamic VM environments compared to monolithic VPN multicast approaches.
3Reliability
If overlay edge nodes process multicast data frames, then multicast delivery is achieved, but bandwidth is wasted when not all nodes need to receive all multicast traffic
Solution Approach 1:
The patent implements feedback mechanisms where overlay edge nodes send queries to the multicast router to determine whether they should receive specific multicast data frames. The multicast router maintains knowledge of multicast group memberships and uses this information to selectively forward frames only to overlay edge nodes that have members interested in that multicast traffic. This feedback loop prevents bandwidth waste by ensuring that multicast frames are delivered only where needed, while still maintaining reliable delivery to all interested recipients.
Data Source
AI summary
A method implemented by a controller in a software defined network (SDN), the method comprising sending, to an overlay edge node, a query message comprising a client specific multicast address, receiving, from the overlay edge node, one or more report messages corresponding to the query message, wherein each of the one or more report messages comprises an address of each of one or more virtual machines (VMs) coupled to the overlay edge node, and updating membership of a multicast group, which is identified by the client specific multicast address, such that the one or more VMs are members in the updated membership of the multicast group.