Distributed ARP Scaling for Packet Forwarding Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current packet forwarding architectures are limited in scalability due to memory constraints, particularly in handling large numbers of connected hosts and virtualization environments, where each packet processor must store complete ARP bindings, limiting the number of hosts that can be served.
Innovation Solution
Implementing a system with distributed ARP scaling across multiple packet processors, where each processor stores only its own ARP bindings and uses relay entries for hosts connected to other processors, along with link aggregation for resilience and load balancing, allowing for next hop scaling and efficient LAG member selection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If each packet processor stores complete ARP bindings for all hosts, then next hop lookup can be performed solely by ingress packet processors, but the number of hosts that can be connected is limited by the memory available to a single packet processor
Solution Approach 1:
The patent divides the ARP binding storage across multiple packet processors, with each processor storing only a portion of the total ARP bindings. This segmentation allows the system to serve more hosts than a single processor could handle, directly resolving the contradiction between ease of ingress lookup and the quantity of hosts that can be served.
Solution Approach 2:
The patent introduces a new dimension to the lookup architecture by adding egress packet processors that perform supplementary lookups. This transforms the system from a single-dimension ingress-only lookup to a two-dimensional ingress-egress collaborative lookup, enabling scalable host support while maintaining operational simplicity.
2Device complexity
If the maximum number of ARP table entries is limited by the ARP binding storage space available to a single packet processor, then adding packet processors does not increase the number of hosts that can be served, but this reduces the scalability of the system
Solution Approach 1:
The patent merges the lookup capabilities of multiple packet processors (both ingress and egress) to collectively handle ARP bindings for a large number of hosts. By combining the storage and processing resources of multiple processors, the system achieves scalability while managing the complexity of ARP table entries across the distributed architecture.
3Adaptability or versatility
If virtualization is used with many virtual hosts per physical host, then the need for scalable packet forwarding device architectures increases, but each virtual host requires a separate ARP binding increasing memory requirements
Solution Approach 1:
The patent applies segmentation by distributing ARP bindings for virtualized hosts across multiple packet processors. Each processor stores a subset of the ARP bindings, allowing the system to support many virtual hosts per physical host without overwhelming the memory of any single processor, thus enabling virtualization scalability.
Data Source
AI summary
The subject matter described herein includes methods, systems, and computer readable media for next hop scaling with link aggregation. According to one aspect of the subject matter described herein, a system for next hop scaling is provided. The system includes a packet forwarding device including a plurality of packet processors for performing next hop and link aggregation group (LAG) selection operations. Within this plurality of packet processors, ingress packet processors are configured to indicate, for received packets that have a next hop on a different packet processor, that an egress next hop selection operation is needed. Egress packet processors of the plurality of packet processors are configured to perform the egress next hop and member selection operations for the packets for which an egress next hop selection operation is indicated, wherein forwarding of the packets is limited to active LAG group members local to the egress packet processor.