ML Workload Placement for Data Center Resource Balancing

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing workload distribution methods in data centers fail to effectively allocate resources to incoming workloads, leading to uneven workload distribution, performance issues, and potential infrastructure outages due to lack of consideration for workload characteristics and resource utilization variations.

Innovation Solution

A computerized system using machine learning to predict future load and rank infrastructure elements based on multidimensional data sets, ensuring optimal resource allocation by selecting the most suitable execution environment for incoming workloads.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If standard load balancers are used to distribute incoming workloads to multiple execution endpoints, then workload distribution is achieved, but the distribution is uneven because all workloads are treated equally without considering their different resource requirements and execution loads

Engineering Contradiction:

Improveworkload distribution efficiencyVSAvoidworkload placement accuracy

Core Design Contradiction:

ProductivityVSManufacturing precision

Solution Approach 1:

The patent applies local quality by treating different infrastructure elements with different characteristics based on their specific properties. The system predicts future load for each infrastructure element individually and uses these predictions to make customized placement decisions, rather than applying a uniform distribution strategy. This allows each workload to be placed on the most suitable infrastructure element based on its specific resource requirements and the predicted state of target elements.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The system dynamically changes the placement strategy parameters based on predicted future load conditions. Instead of using static load balancing rules, the patent continuously updates placement decisions based on real-time predictions of infrastructure element states, adjusting the allocation strategy to match changing system conditions and workload characteristics.

Inventive Principle:

Principle #35Parameter changes

2Speed

If workload placement decisions are made without considering future load predictions, then placement speed is maintained, but resource utilization becomes suboptimal and performance issues arise

Engineering Contradiction:

Improveworkload placement speedVSAvoidresource utilization efficiency

Core Design Contradiction:

SpeedVSProductivity

Solution Approach 1:

The system performs preliminary actions by predicting the future load state of infrastructure elements before making placement decisions. The load prediction mechanism proactively assesses upcoming resource availability and identifies suitable placement targets in advance, allowing the system to make informed decisions that optimize both speed and resource utilization without waiting for actual load conditions to manifest.

Inventive Principle:

Principle #10Preliminary action

3Ease of operation

If infrastructure elements are selected without considering workload characteristics, then placement simplicity is maintained, but performance issues and application behavior anomalies occur

Engineering Contradiction:

Improveplacement decision simplicityVSAvoidapplication performance stability

Core Design Contradiction:

Ease of operationVSReliability

Solution Approach 1:

The system enables self-service by allowing workload characteristics to automatically guide their own placement decisions. The load predictor and placement decider work autonomously to match workload requirements with suitable infrastructure elements based on predicted future states, eliminating the need for manual intervention while ensuring performance stability through intelligent, characteristic-aware placement.

Inventive Principle:

Principle #25Self-service

4Device complexity

If existing load balancing methods are used without machine learning predictions, then system complexity is reduced, but workload distribution becomes uneven and scalability is impaired

Engineering Contradiction:

Improveload balancing system complexityVSAvoiddata center scalability

Core Design Contradiction:

Device complexityVSProductivity

Solution Approach 1:

The patent replaces traditional mechanical load balancing systems with an intelligent system based on machine learning predictions. Instead of relying on simple algorithms like round-robin or static weighted distribution, the system uses predictive analytics to dynamically determine optimal placement, substituting complex computational intelligence for simpler mechanical distribution methods to achieve better scalability and performance.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS11941450B2Automatic placement decisions for running incoming workloads on a datacenter infrastructure

Publication Date: 2024.03.26 DELL PROD LP

AI summary

A system and method place an incoming workload within a data center having infrastructure elements (IEs) for execution. Instrumentation data are collected for both individual IEs in the data center, and workload instances executing on each of these IEs. These data are used to train a future load model according to machine learning techniques, especially supervised learning. Future loads, in turn, are used to train a ranking model that ranks IEs according to suitability to execute additional workloads. After receiving an incoming workload, the first model is used to predict, for each IE, the load on its computing resources if the workload were executed on that IE. The resulting predicted loads are then fed into the second model to predict the best ranking of IEs, and the workload is placed on the highest-ranked IE that is available to execute the workload.