In-Network Multicasting of Load Statistics Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional IP multicast solutions face scalability limitations and introduce overheads and high latencies, making them unsuitable for modern cloud datacenter environments that require high throughput and low latency for one-to-many communications.
Innovation Solution
A software-defined network approach where an underlay-network controller generates and distributes match-action rules and special multicast headers in SDN networks, enabling efficient configuration of multicast groups and packet forwarding without relying on overlay traffic, and includes multicast tree information directly in packet headers to eliminate the need for large forwarding tables.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional IP multicast solutions are used, then multicast communication is enabled, but scalability is limited and cannot support hundreds of thousands of tenants
Solution Approach 1:
The patent embeds multicast tree information directly in packet headers (adding a dimension to the packet structure), allowing switches to forward packets based on header information rather than maintaining large multicast tables. This transforms the scaling problem from table-size limitations to header-size limitations, enabling support for hundreds of thousands of tenants.
Solution Approach 2:
The patent creates compact copies of multicast tree information and embeds them in packet headers. Instead of storing complete multicast state in switch tables, only essential routing information is copied into the packet header, dramatically reducing the state that switches must maintain while preserving forwarding capability.
2Adaptability or versatility
If overlay IP multicast is implemented, then multicast support is provided for cloud environments, but scalability remains limited
Solution Approach 1:
The patent extracts multicast tree information from the control plane and embeds it directly in data plane packets. This separation removes the burden of maintaining large multicast state from the control plane and switch memory, allowing the system to scale to cloud environment requirements without proportional increases in control plane overhead.
Solution Approach 2:
The packet headers carry their own routing information (multicast tree data), making packets self-describing. Switches don't need to query control plane or maintain external tables to determine forwarding behavior; the packet itself contains all necessary information, enabling autonomous forwarding decisions at scale.
3Device complexity
If unicast-based solutions are used, then implementation is simple, but overheads and high latencies negatively impact network throughput
Solution Approach 1:
The patent segments multicast tree information into compact representations that fit in packet headers. This segmentation allows the system to use simple header-based forwarding (similar to unicast simplicity) while maintaining efficient multicast group forwarding, achieving both implementation simplicity and high network throughput.
4Adaptability or versatility
If PIM/IGMP multicast implementations are used in switches, then multicast routing is enabled, but switching chips cannot support hundreds of thousands of multicast groups
Solution Approach 1:
The patent moves multicast state from switch memory (two-dimensional table lookup) to packet headers (one-dimensional data structure). This dimensional shift allows the system to support hundreds of thousands of multicast groups because the limitation transitions from finite switch memory to practically unlimited header space combined with distributed storage across the network.
Solution Approach 2:
The patent creates minimal copies of routing information embedded in packet headers rather than maintaining complete multicast state in switch tables. This copying approach reduces the quantity of data that must be stored in switching chip memory from millions of entries to compact header structures, enabling support for vastly larger numbers of multicast groups.
5Device complexity
If bloom filters are used to encode multicast distribution trees, then multicast routing information is compacted, but significant time latencies are introduced
Solution Approach 1:
The patent changes the parameter representation from probabilistic bloom filter structures to deterministic header fields that can be directly parsed and matched. This parameter change eliminates the computational overhead of bloom filter operations while maintaining compact header sizes, achieving both space efficiency and low processing latency.
Data Source
AI summary
In an embodiment, a computer-implemented method for highly-scalable, in-network multicasting of statistics data is disclosed. In an embodiment, a method comprises: receiving, from an underlay controller, a match-and-action table that is indexed using one or more multicast (“MC”) group identifiers and includes one or more special MC headers; detecting a packet carrying statistics data; determining whether the packet includes an MC group identifier; in response to determining that the packet includes the MC group identifier: using the MC group identifier, retrieving a special MC header, of the one or more special MC headers, from the match-and-action table; generating an encapsulated packet by encapsulating the packet with the special MC header; and providing the encapsulated packet to an interface controller for transmitting the encapsulated packet to one or more physical switches.