Bin Packing Algorithm for Data Center Node Provisioning

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing data center node provisioning methods, such as bin packing, do not effectively consider performance degradation when placing large-scale virtual execution environments, leading to resource overconsumption and decreased node performance.

Innovation Solution

Implement a bin packing algorithm that takes into account a degradation metric, measuring real-time resource availability and sensitivity, to select nodes less sensitive to additional loads, ensuring efficient resource allocation and maintaining performance by avoiding nodes with excessive degradation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If traditional bin packing algorithm is used to provision virtual execution environments, then resource utilization is improved, but node performance degradation occurs due to overconsumption

Engineering Contradiction:

Improveresource utilizationVSAvoidnode performance

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The system implements feedback by continuously monitoring node performance metrics and degradation levels. The bin packing algorithm receives real-time feedback about node performance states and adjusts provisioning decisions accordingly, preventing placement on nodes that would exceed acceptable degradation thresholds. This closed-loop control resolves the contradiction by dynamically balancing resource utilization against performance maintenance.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The provisioning system transitions from static bin packing to dynamic decision-making by considering real-time node performance characteristics. The system adapts provisioning strategies based on current node states, varying the placement decisions according to measured degradation metrics. This dynamic approach allows optimal resource utilization while preventing performance degradation through context-aware provisioning.

Inventive Principle:

Principle #15Dynamics

2Productivity

If large-scale virtual execution environments are provisioned without considering degradation metrics, then service capacity is improved, but resource overconsumption occurs

Engineering Contradiction:

Improveservice capacityVSAvoidresource overconsumption

Core Design Contradiction:

ProductivityVSLoss of energy

Solution Approach 1:

The system uses feedback mechanisms to monitor actual resource consumption and performance degradation resulting from provisioning decisions. By measuring the impact of large-scale virtual environment deployments on host node resources, the system adjusts subsequent provisioning to prevent overconsumption while maintaining service capacity targets.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system implements a threshold-based approach where nodes exceeding degradation thresholds are temporarily excluded from provisioning candidates. This allows the system to provision large-scale environments on suitable nodes while avoiding resource overconsumption on degraded nodes, effectively managing resource lifecycles based on performance conditions.

Inventive Principle:

Principle #27Cheap short-living objects (Disposable)

3Productivity

If nodes with high resource availability are selected for provisioning, then resource allocation efficiency is improved, but node sensitivity to additional load increases

Engineering Contradiction:

Improveresource allocation efficiencyVSAvoidnode sensitivity to load

Core Design Contradiction:

ProductivityVSObject-affected harmful factors

Solution Approach 1:

The system incorporates feedback about node sensitivity and degradation metrics into the bin packing decision process. Rather than solely relying on static resource availability metrics, the system uses real-time feedback about how nodes respond to additional loads to make informed provisioning decisions that balance allocation efficiency with load sensitivity considerations.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system changes the parameters used for node selection from static resource availability to dynamic metrics including degradation levels and load sensitivity. By transforming the selection criteria to incorporate these additional parameters, the system achieves better balance between allocation efficiency and sensitivity management through multi-dimensional evaluation.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS10469616B2Data driven bin packing implementation for data centers with variable node capabilities

Publication Date: 2019.11.05 RED HAT INC

AI summary

A method includes receiving, by a processing device executing a scheduler, a first request to provision a software application, the request comprising a resource requirement associated with executing the software application, requesting, from a node of a data center, a performance degradation index value of the node, the performance degradation index value reflecting performance degradation of at least one of a processor resource, a memory resource, or a network resource, determining, in view of the performance degradation index values, whether the node meets the resource requirement associated with executing the software application, and responsive to determining that the node meets the resource requirement, provisioning the software application to execute on the node.