Pinning Bi-directional Network Traffic to Service Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional hash-based equal-cost multipath routing and load balancing methods fail to ensure that bi-directional network traffic between hosts is distributed to the same service device, preventing complete state maintenance for network connections and hindering stateful processing.
Innovation Solution
Creating load balancer groups for each domain, enabling symmetric hashing, and synchronizing hash tables across these groups to guarantee that both forward and reverse traffic between hosts is routed to the same service device, ensuring bi-directional state information is maintained.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional hash-based equal-cost multipath routing is used to distribute network traffic, then load balancing is achieved, but bi-directional traffic between hosts cannot be ensured to reach the same service device
Solution Approach 1:
The patent applies asymmetry by introducing a domain identifier into the hash function, creating asymmetric hash tables that are specific to each service device's domain. This allows the system to maintain symmetry in traffic distribution (both directions go to same device) while using asymmetric hash table structures tailored to each device's domain, resolving the contradiction between load balancing and state maintenance.
Solution Approach 2:
The patent segments the load balancing mechanism by creating separate hash tables for each service device, where each hash table is associated with a specific domain. This segmentation allows independent management of traffic routing per device while ensuring that bi-directional traffic within a domain is consistently routed to the same device, maintaining both productivity and reliability.
2Reliability
If service devices maintain bi-directional state information per network connection, then stateful processing is enabled, but traffic must be consistently routed to the same device which complicates load balancing
Solution Approach 1:
The patent introduces domain identifiers as an intermediary element in the routing decision process. By hashing the domain identifier along with source and destination addresses, the system creates a deterministic routing key that consistently directs bi-directional traffic to the same service device without requiring complex routing logic, thus enabling stateful processing while managing complexity.
Solution Approach 2:
The patent changes the parameters used in hash-based routing by incorporating domain identifiers into the hash function. This parameter change transforms the routing behavior to ensure that traffic within the same domain is consistently routed to the same service device, enabling state maintenance while keeping the routing mechanism relatively simple through standard hashing operations.
3Reliability
If hash tables are synchronized across service devices with common buckets mapping to single devices, then bi-directional traffic pinning is achieved, but system configuration complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-configuring domain-specific hash tables on each service device before traffic routing begins. Each device's hash table is pre-populated with domain identifiers and their corresponding routing information, allowing deterministic routing decisions without real-time synchronization complexity, thus achieving traffic consistency while managing configuration complexity.
Data Source
AI summary
Techniques for ensuring that, in the context of network traffic load-balanced across a plurality of service devices connected to a network device, all of the bi-directional traffic between a given pair of hosts residing in different domains is sent to the same service device, where a “domain” is a group of one or more hosts/subnets that is reachable by a service device via an interface of that device. In one set of embodiments, these techniques can include (1) creating a load balancer group on the network device for each domain defined on the service devices, such that the load balancer group for a given domain D includes all of the service device interfaces mapped to D, (2) enabling symmetric hashing with respect to each load balancer group, and (3) synchronizing the hash tables of the load balancer groups such that a given hash bucket (across all hash tables) maps to an interface of a single service device.