Hierarchical Overlay Network Replication Across Datacenters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current network systems face limitations in supporting a large number of multicast groups across physical networks, especially in datacenters and provider networks, due to infrastructure constraints and the complexity of managing IP multicast, which hinders network virtualization and increases bandwidth usage.
Innovation Solution
A hierarchical method for scalable replication of broadcast, unknown unicast, and multicast messages across logical overlay networks spanning multiple physical underlay networks, using tunnel endpoints and proxy endpoints to replicate messages within and across network segments and physical networks, even when physical networks do not fully support multicast protocols.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If IP multicast is enabled in physical networks to support multicast groups, then multicast delivery capability is improved, but device complexity and management complexity increase significantly
Solution Approach 1:
The patent introduces a control plane as an intermediary that manages multicast group information centrally. The control plane receives join requests, determines optimal proxy endpoints, and distributes group membership information to tunnel endpoints. This mediator approach allows multicast functionality without requiring complex distributed IGMP/PIM protocols across all network devices, thus improving multicast capability while reducing management complexity.
Solution Approach 2:
The patent uses proxy endpoints that maintain copies of multicast group membership information. Instead of requiring every network device to independently manage multicast state, selected proxy endpoints store and replicate group membership data. This copying mechanism enables multicast delivery while simplifying the complexity burden on individual devices, as only designated proxies need to maintain state information.
2Productivity
If a large number of multicast groups are supported across datacenters, then network virtualization scalability is improved, but bandwidth usage increases due to redundant traffic replication
Solution Approach 1:
The patent segments the network into multiple datacenters with designated proxy endpoints in each. Instead of replicating multicast traffic across all possible paths, the system divides the network hierarchy and routes traffic through segment boundaries at proxy endpoints. This segmentation allows scalable support for many multicast groups while optimizing bandwidth usage by avoiding redundant transmissions across segment boundaries.
Solution Approach 2:
The patent introduces a hierarchical dimension to multicast routing by adding proxy endpoints at multiple levels (intra-datacenter and inter-datacenter). This dimensional approach allows traffic to be replicated efficiently within segments and then consolidated at higher hierarchical levels, reducing overall bandwidth consumption compared to flat replication across all nodes.
3Productivity
If physical network multicast capabilities are leveraged, then replication efficiency is improved, but infrastructure requirements and compatibility constraints increase
Solution Approach 1:
The control plane acts as an intermediary that abstracts the complexity of physical network multicast capabilities. It determines which proxy endpoints should leverage hardware multicast support and manages the distribution of multicast traffic accordingly. This mediator approach allows the system to exploit available hardware capabilities when present while maintaining compatibility with networks that have limited or no native multicast support.
Solution Approach 2:
The patent dynamically adjusts routing parameters based on the multicast capabilities of the physical network. The control plane can modify traffic distribution strategies, proxy endpoint selections, and replication behavior according to the detected hardware capabilities. This parameter adaptation allows efficient use of hardware multicast when available while gracefully degrading to software-based replication on networks with limited capabilities.
Data Source
AI summary
Some embodiments provide a method of replicating messages for a logical network. At a particular tunnel endpoint in a particular datacenter, the method receives a message to be replicated to members of a replication group. The method replicates the message to a set of tunnel endpoints of the replication group located in a same segment of the particular datacenter as the particular tunnel endpoint. The method replicates the message to a first set of proxy endpoints of the replication group, each of which is located in a different segment of the particular datacenter and for replicating the message to tunnel endpoints located in its respective segment of the particular datacenter. The method replicates the message to a second set of proxy endpoints of the replication group, each of which is located in a different datacenter and for replicating the message to tunnel endpoints located in its respective datacenter.