Server Selection in Highly Available Networks Without Load Balancers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Highly available networks rely on Server Load Balancers (SLBs) to reroute traffic in case of server failures, but this setup can be vulnerable if the load balancer itself becomes inoperable, leading to service unavailability.
Innovation Solution
Implementing a networked computing system where network devices and management application instances use a locally stored registry to select and communicate with management application instances without relying on a load balancer, utilizing selection logic and UI logic to route requests directly to active servers, and adjusting priorities based on response times and availability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a Server Load Balancer is used to reroute traffic in case of server failures, then service availability is improved, but the system introduces a single point of failure that can lead to service unavailability
Solution Approach 1:
The patent extracts the load balancing functionality from a centralized SLB device and distributes it to individual network devices. Each network device maintains its own load balancing logic and registry of management application instances, eliminating the dependency on a single external load balancer while preserving traffic routing capabilities
Solution Approach 2:
The centralized load balancing function is segmented and distributed across multiple network devices. Instead of one SLB managing all traffic, each network device independently performs load balancing decisions based on its local registry, creating multiple independent decision points that eliminate the single point of failure
2Reliability
If network devices use a locally stored registry to select management application instances directly, then dependency on load balancer is reduced, but network device complexity increases
Solution Approach 1:
Network devices perform preliminary actions by maintaining local registries of management application instances and their health statuses. This pre-established information allows devices to make immediate load balancing decisions without querying a central SLB, enabling direct communication with available instances while keeping added complexity minimal
Data Source
AI summary
Systems, methods, and other embodiments associated with a highly available network computer system that does not utilize a load balancer are described. In one embodiment, the system includes a plurality of management application instances configured to manage network devices and a plurality of network devices. Each network device includes computer storage medium storing a registry of management application instances. The registry records, for each management application instance, connection information used to communicate with the management application instance and an associated priority for each management application instance. The network device also includes an agent and selection logic. The agent is configured to identify a network communication intended for a management application instance. The selection logic is configured to access the registry; identify a management application instance in the registry having a highest associated priority; and cause the agent to transmit the network communication to the highest priority management application instance.