Distributed Next Hop Scaling for Packet Forwarding Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current packet forwarding devices are limited in scalability due to memory constraints, as each packet processor must store a complete set of ARP bindings for all hosts, restricting the number of connected hosts and hindering performance, especially in virtualization environments where many virtual hosts require separate bindings.
Innovation Solution
Implementing a system with distributed next hop scaling, where ARP bindings are stored across multiple packet processors, with each processor responsible only for its directly connected hosts, using relay entries for hosts connected to other processors, allowing for efficient packet forwarding and increased memory utilization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If each packet processor stores a complete set of ARP bindings for all hosts, then any packet processor can forward packets to any host, but the number of hosts is limited by the memory of a single packet processor
Solution Approach 1:
The patent divides the ARP binding storage responsibility among multiple packet processors based on their connected hosts. Each packet processor stores ARP bindings only for hosts directly connected to it, rather than storing complete ARP tables for all hosts. This segmentation allows the system to serve more hosts by distributing the storage burden across multiple processors.
Solution Approach 2:
The patent introduces a new dimension to packet forwarding by adding a relay mechanism. When a packet processor receives a packet for a host not directly connected to it, it relays the packet to the appropriate packet processor that has the destination host connected. This adds a forwarding dimension that eliminates the need for each processor to have complete ARP tables.
2Quantity of substance
If the maximum number of ARP table entries is limited by the memory of a single packet processor, then memory constraints are satisfied, but adding packet processors does not increase the number of hosts that can be served
Solution Approach 1:
The patent segments the ARP binding storage function across multiple packet processors, where each processor maintains bindings for its locally connected hosts. This segmentation enables linear scaling of host capacity with the number of packet processors, as each additional processor contributes its local host bindings to the overall system capacity.
Solution Approach 2:
The patent introduces relay entries as intermediaries that enable packet processors to forward packets to other packet processors. These relay entries contain the egress port information needed to route packets to the correct destination processor, acting as a mediator that connects the distributed ARP binding storage architecture.
3Adaptability or versatility
If each packet processor stores complete ARP bindings, then full-mesh traffic forwarding is enabled, but scalability is reduced by a factor of 1/(number of packet processors)
Solution Approach 1:
The patent segments the complete ARP binding set into multiple distributed bindings, with each packet processor storing bindings for its locally connected hosts. This segmentation maintains full-mesh traffic capability through relay forwarding while improving scalability by eliminating the need for redundant storage of complete ARP tables at each processor.
Solution Approach 2:
The patent adds a relay forwarding dimension to the traditional direct forwarding model. Instead of requiring each processor to directly know all hosts (zero-dimensional complete knowledge), the system uses relay entries that provide one-dimensional path information to reach distant hosts, reducing memory requirements while maintaining connectivity.
Data Source
AI summary
The subject matter described herein includes a packet forwarding device that implements next hop scaling. Rather than storing a complete set of next hop bindings at each packet processor, the storage of next hop bindings is distributed among packet processors in the packet forwarding device such that each packet processor stores next hop bindings for the hosts that are directly connected to the packet processor. For hosts that are not directly connected to a packet processor, the packet processor stores relay entries. Because of the distributed storage of next hop bindings, the number of hosts that can be served by a single packet forwarding device is increased over packet forwarding devices where each packet processor stores a complete set of next hop bindings for all connected hosts.