Container Cluster Rebalancing Using Predicted Worker Node Migration

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing containerized application clusters face challenges in dynamically rebalancing worker nodes to handle varying workloads and resource demands, leading to potential service quality degradation due to resource constraints.

Innovation Solution

A computer-implemented method and system that utilizes a machine learning model to predict future workload and dynamically rebalance worker nodes by requesting and migrating nodes from peer clusters based on factors like distance, energy, and hardware prerequisites, using container orchestration techniques.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If worker nodes are statically allocated to containerized application clusters, then device complexity is reduced and ease of operation is improved, but adaptability to varying workload demands deteriorates and service quality degradation occurs

Engineering Contradiction:

Improveadaptability to workload fluctuationsVSAvoidcomplexity of dynamic rebalancing system

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent implements dynamic worker node allocation where the orchestration system continuously monitors workload demands and automatically migrates worker nodes between clusters based on real-time needs. This transforms the static allocation model into a dynamic one, allowing clusters to adapt their resource capacity flexibly without manual intervention, thereby resolving the contradiction between adaptability and system complexity.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system employs feedback mechanisms through workload monitoring and demand prediction algorithms that continuously assess cluster states and trigger rebalancing actions. The orchestration system receives feedback about workload patterns, predicts future demands, and automatically adjusts worker node distribution accordingly, enabling adaptive resource allocation while maintaining manageable system complexity through automated control loops.

Inventive Principle:

Principle #23Feedback

2Productivity

If worker nodes are migrated between peer clusters dynamically, then adaptability to workload demands is improved, but loss of time occurs during migration processes

Engineering Contradiction:

Improveworkload handling efficiencyVSAvoidtime for worker node migration

Core Design Contradiction:

ProductivityVSLoss of time

Solution Approach 1:

The system performs preliminary actions by pre-positioning worker nodes in clusters where they are most likely needed based on predicted workload patterns. The orchestration system analyzes historical and real-time workload data to anticipate future demands, proactively migrating worker nodes before peak loads occur, thereby minimizing service disruption and reducing the effective migration time impact on productivity.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent implements optimized migration processes that accelerate worker node transitions between clusters by parallelizing migration tasks, pre-establishing network connections, and minimizing state transfer overhead. The system rushes through critical migration phases by prioritizing essential workload migration over non-critical data synchronization, thereby reducing overall migration time while maintaining workload handling efficiency.

Inventive Principle:

Principle #21Skipping (Rushing through)

3Productivity

If worker nodes are migrated from peer clusters, then resource allocation efficiency is improved, but reliability of peer clusters deteriorates due to resource depletion

Engineering Contradiction:

Improveresource allocation efficiencyVSAvoidreliability of peer clusters

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The orchestration system employs feedback control by continuously monitoring resource levels in both source and destination clusters during migration processes. When a peer cluster's resources fall below threshold levels, the system automatically adjusts migration rates or halts further migrations from that cluster, preventing resource depletion and maintaining peer cluster reliability while still achieving overall resource allocation efficiency through coordinated rebalancing.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system implements beforehand cushioning by maintaining reserve worker nodes and establishing minimum resource threshold policies for peer clusters before migration operations begin. These protective measures ensure that even during active rebalancing, peer clusters retain sufficient resources to maintain their reliability and service levels, preventing resource depletion while enabling efficient resource reallocation across the federated system.

Inventive Principle:

Principle #11Beforehand cushioning (Prior cushioning)

Data Source

PatentUS12609985B2Dynamic rebalancing of containerized application cluster

Publication Date: 2026.04.21 INTERNATIONAL BUSINESS MACHINE CORPORATION

AI summary

A moving server containerized application cluster can be dynamically rebalanced. Future incoming workload to a first moving server containerized application cluster is determined. Based on the future incoming workload, at least one worker node is requested from a peer moving server containerized application cluster. At least one worker node is caused to migrate from the peer moving server containerized application cluster to the first moving server containerized application cluster.