Server Cluster Load Balancing via Virtual Server Neighbors

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing server clusters face challenges in achieving high availability and scalability while managing session contexts, particularly in network applications, due to bottlenecks in session management and load balancing, especially when handling failures and high-throughput scenarios.

Innovation Solution

The solution involves a scheduler that uses a hash-based lookup method to distribute requests across multiple virtual servers, allowing for efficient load balancing and failover without requiring expensive storage area networks or high network bandwidth, by defining neighboring virtual servers to take over sessions in case of failures and using meta segments to reduce memory requirements.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If session information is stored in the scheduler to enable load balancing for network applications with contexts, then load balancing capability is improved, but the scheduler becomes a bottleneck and memory requirements increase

Engineering Contradiction:
Improveload balancing capabilityVSAvoidsystem throughput
Core Design Contradiction:
Adaptability or versatilityVSProductivity

Solution Approach 1:

The patent extracts session information management from the scheduler by implementing it within virtual server instances themselves. Each virtual server maintains its own session information locally, eliminating the need for the scheduler to store and manage session data, thus removing the scheduler as a bottleneck while enabling load balancing for applications with contexts

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

Session information is stored locally within each virtual server instance rather than centrally in the scheduler. This local storage approach allows each virtual server to independently access and manage its own session data, improving system throughput by eliminating centralized access bottlenecks while maintaining load balancing capabilities

Inventive Principle:
Principle #3Local quality

2Adaptability or versatility

If session information is duplicated across all servers in the cluster, then load balancing is improved, but network bandwidth requirements and system complexity increase

Engineering Contradiction:
Improveload balancing capabilityVSAvoidnetwork bandwidth consumption
Core Design Contradiction:
Adaptability or versatilityVSQuantity of substance

Solution Approach 1:

The patent extracts session information from centralized or distributed storage and embeds it directly within each virtual server instance. This extraction eliminates the need for network-based session information sharing, reducing network bandwidth consumption while maintaining load balancing capability through local session access

Inventive Principle:
Principle #2Taking out (Extraction)

3Adaptability or versatility

If a shared storage area network is used to store session information, then session accessibility is improved, but system cost and complexity increase

Engineering Contradiction:
Improvesession accessibilityVSAvoidstorage infrastructure complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent extracts session information management from external shared storage infrastructure and implements it within the virtual server instances themselves. This eliminates the requirement for expensive shared storage area networks while maintaining session accessibility through local storage within each virtual server

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

Each virtual server instance maintains its own copy of session information locally, eliminating the need for shared storage infrastructure. This copying approach within virtual memory spaces provides session accessibility without the complexity and cost of external storage systems

Inventive Principle:
Principle #26Copying

4Ease of operation

If the scheduler maintains a table of sessions and provides lookup functionality, then session management is improved, but the scheduler becomes a performance bottleneck

Engineering Contradiction:
Improvesession management capabilityVSAvoidscheduler processing speed
Core Design Contradiction:
Ease of operationVSProductivity

Solution Approach 1:

The patent extracts session management functionality from the scheduler and implements it within each virtual server instance. Each virtual server independently manages its own session table and provides lookup functionality, eliminating the scheduler as a performance bottleneck while maintaining ease of session management

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

Each virtual server instance independently manages its own session information without requiring scheduler intervention. The virtual servers perform self-service session management, including storage, retrieval, and updates, which eliminates the scheduler as a bottleneck while maintaining session management capability

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP2583417B1Server cluster
Publication Date: 2018.08.29 NOKIA SOLUTIONS & NETWORKS OY
  • EP2583417B1 patent drawingFigure 1~2
  • EP2583417B1 patent drawingFigure 3~4
  • EP2583417B1 patent drawingFigure 5~6

AI summary

A server cluster is described, which enables load balancing between servers in the cluster. At least some of the servers in the cluster are divided into a plurality of virtual servers, wherein each virtual server is associated with a neighbouring server, which neighbouring server acts as a backup for that virtual server. The neighbouring server of each virtual server of a particular server is part of a different physical server to the virtual server, such that in the event that a physical server is unavailable for use, the load of the virtual servers of that physical server is split between a number of different physical servers, thereby reducing the likelihood of overloading any particular physical server.