Selective Multicast Replication for Load Balancing Node Changes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Load balancing clusters face challenges in reliably mapping client sessions to nodes due to changes such as failovers, scaling, and denial of service attacks, which existing methods address inadequately by requiring modifications to the IP stack.
Innovation Solution
The method involves selective multicast replication of data packets, using a hashing function to determine next-hop table indices and forwarding information, allowing for graceful node removal and hitless node addition without modifying the IP stack, by employing a multicast replication status to direct packets through tertiary, secondary, or primary forwarding groups.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If selective multicast replication of data packets is implemented, then load balancing reliability is improved during node changes, but device complexity increases due to additional forwarding groups and lookup tables
Solution Approach 1:
The patent segments the forwarding logic into multiple forwarding groups (primary, secondary, tertiary) and uses a group lookup table to select among them based on multicast replication status. This segmentation allows the system to maintain multiple forwarding paths without requiring complex decision-making at each packet processing stage, thereby improving reliability while managing device complexity through structured organization.
Solution Approach 2:
The patent introduces a group lookup table as an intermediary component that sits between the packet processing logic and the actual forwarding groups. This intermediary translates packet information into appropriate forwarding group selections, abstracting the complexity of multiple forwarding paths and enabling reliable load balancing during node changes without directly increasing the complexity of core packet processing logic.
2Adaptability or versatility
If multicast replication status is used to direct packets through different forwarding groups, then adaptability to node changes is improved, but processing time increases due to additional lookup operations
Solution Approach 1:
The patent performs preliminary actions by pre-configuring multiple forwarding groups (primary, secondary, tertiary) and pre-computing the group lookup table mappings before node changes occur. When node failures or scaling events happen, the system can immediately switch to appropriate forwarding groups without performing complex real-time calculations, thus maintaining high adaptability while minimizing additional processing time during dynamic events.
Solution Approach 2:
The patent applies different forwarding group qualities locally based on the specific node change scenario. Instead of using a single uniform forwarding path for all packets, the system selectively directs packets through primary, secondary, or tertiary forwarding groups based on the multicast replication status and specific node conditions. This local quality approach allows the system to adapt to node changes efficiently while minimizing processing overhead by only performing additional lookups when necessary.
3Productivity
If hashing function is used to determine next-hop table indices, then load distribution is improved, but measurement precision requirements increase for hash key generation
Solution Approach 1:
The patent implements a universal hashing function that serves multiple purposes: it generates hash keys for load distribution, determines next-hop table indices for efficient packet routing, and works across different forwarding groups (primary, secondary, tertiary). This multi-functional hashing approach improves load distribution efficiency while managing precision requirements by using the same hash key generation logic across different scenarios, reducing the need for separate precision-critical measurement systems.
Data Source
AI summary
Load balancing through selective multicast replication of data packets in a computer network. Specifically, selective multicast replication entails facilitating the graceful removal or hitless addition of one or more load balancing nodes in a load balancing cluster by multicasting, rather than encapsulating, received data packets from any given client to the load balancing cluster. Further, selective multicast replication considers whether received data packets are synchronize (SYN) or non-synchronize (non-SYN) data packets, in order to employ an appropriate forwarding group towards forwarding the received data packets to one or more appropriate load balancing nodes in the load balancing cluster.


