Multicast Proxy for Non-Tree Data Center Topologies
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Multicast group membership discovery protocols such as IGMP and MLD fail to operate correctly in networks without tree topologies, as not all routers can receive traffic from upstream sources, leading to issues in IPv4 and IPv6 networks.
Innovation Solution
Implementing a multicast group membership discovery protocol proxy on compute servers in data center networks that can run virtual machines, which receives report messages, determines if the multicast data source is hosted locally, and forwards requests accordingly to ensure data delivery within the network.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If IGMP/MLD protocols are deployed in networks without tree topology, then the protocols fail to operate correctly because not all routers can receive traffic from upstream sources
Solution Approach 1:
The patent introduces an intermediary mechanism where the virtualization layer (hypervisor) acts as a mediator between the multicast protocol and the underlying network infrastructure. The hypervisor intercepts multicast report messages from virtual machines, determines the appropriate physical network interface for forwarding based on data source location, and performs the actual forwarding. This intermediary layer abstracts the complexity of non-tree topology handling from the protocol itself, enabling correct operation in arbitrary network topologies.
Solution Approach 2:
The patent segments the multicast protocol functionality into two distinct layers: the virtual machine layer (guest OS) and the physical host layer (hypervisor). The VM generates standard IGMP/MLD report messages as if in a traditional network, while the hypervisor layer handles the complex topology-aware forwarding decisions. This segmentation allows the protocol to remain simple and standardized at the VM level while adapting to complex topologies at the host level.
2Ease of operation
If traditional router election mechanisms are used in non-tree topologies, then traffic forwarding fails because elected routers may not have upstream access to data sources
Solution Approach 1:
The patent implements a feedback mechanism where the hypervisor monitors the location of multicast data sources and uses this information to make intelligent forwarding decisions. When a VM sends a report message, the hypervisor receives feedback about the data source location (through metadata or network monitoring) and adjusts the forwarding path accordingly. This feedback loop ensures that traffic is always routed through interfaces that have actual access to the data source, rather than relying on static router election rules.
Solution Approach 2:
The patent makes the multicast forwarding behavior dynamic rather than static. Instead of relying on fixed router election rules that assume tree topology, the hypervisor dynamically determines the appropriate forwarding interface based on the current location of data sources. The forwarding behavior adapts in real-time to changes in data source location, network topology, or VM migration, ensuring reliable delivery in non-tree topologies.
Data Source
AI summary
Measures, including methods, systems and non-transitory computer-readable storage mediums, for use in processing multicast group membership discovery protocol messages in a data center network including a plurality of compute servers. Each compute server in the plurality is capable of running one or more virtual machines. At a multicast group membership discovery protocol proxy located on a compute server in the plurality, a multicast group membership discovery protocol report message is received requesting delivery of data associated with a specified multicast group from a specified multicast data source. It is determined whether the specified multicast data source is hosted by a virtual machine running on the compute server. In response to a positive determination, a request corresponding to the received multicast group membership discovery protocol report message is forwarded to the virtual machine hosting the specified multicast data source on the compute server.