Breadth-first resource allocation for server fault tolerance

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing resource allocation policies in grid computing are vulnerable to individual server failures, leading to potential total loss of service for applications if a server goes down, as they often concentrate CPU instances on a single server, making the system intolerant to failures.

Innovation Solution

A breadth-first resource allocation policy that distributes application instances evenly across multiple servers, ensuring that if one server fails, the remaining instances can be automatically routed to new servers, maintaining service continuity.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If CPU slots are allocated from a single server until all slots are used, then allocation speed is improved, but system reliability deteriorates because all instances may be concentrated on one server

Engineering Contradiction:

Improveallocation speedVSAvoidsystem reliability

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The patent segments the allocation process into two distinct phases: a breadth-first phase that distributes instances across multiple servers to ensure reliability, and then a depth-first phase that fills remaining capacity. This segmentation resolves the contradiction by ensuring instances are not concentrated on a single server while maintaining efficient allocation.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary distribution of instances across multiple servers before filling individual server capacity. By pre-distributing instances to multiple servers in the breadth-first phase, the system ensures reliability is established before optimization for capacity utilization occurs.

Inventive Principle:

Principle #10Preliminary action

2Reliability

If instances are distributed across multiple servers, then fault tolerance is improved, but allocation complexity increases

Engineering Contradiction:

Improvefault toleranceVSAvoidallocation complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent implements periodic action through its two-phase allocation cycle: first executing breadth-first distribution across servers, then executing depth-first filling of capacity. This periodic alternation between distribution and optimization phases provides a structured, manageable approach to multi-server allocation that reduces complexity while maintaining fault tolerance.

Inventive Principle:

Principle #19Periodic action

3Ease of operation

If servers are selected in order from resource group, then allocation simplicity is improved, but load balancing deteriorates

Engineering Contradiction:

Improveallocation simplicityVSAvoidload balancing

Core Design Contradiction:

Ease of operationVSProductivity

Solution Approach 1:

The patent introduces dynamics by making the server selection process adaptive rather than static. The breadth-first phase dynamically selects servers from the resource group to distribute instances, and the depth-first phase dynamically fills capacity based on remaining availability. This dynamic approach maintains simplicity while achieving effective load balancing.

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS9882973B2Breadth-first resource allocation system and methods

Publication Date: 2018.01.30 JPMORGAN CHASE BANK NA

AI summary

A computing environment adapted with a resource allocation policy, the resource allocation policy being configured to evenly distribute application instances to as many servers as possible.