Virtual Machine Noisy Neighbor Detection Using Hardware Counters

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing cloud infrastructure systems struggle to effectively detect and mitigate noisy neighbors in multi-tenant virtualized environments, leading to resource starvation and performance issues among virtual machines due to heterogeneous workloads that are not initially known to the hypervisor or fleet manager.

Innovation Solution

A method utilizing hardware performance measurement components to determine deviations in utilization through short-term and long-term averages, combined with linear regression, to identify noisy neighbors and implement real-time mitigation strategies.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If multi-tenant virtualized cloud infrastructure is used to share resources among tenants, then resource utilization and cost efficiency are improved, but noisy neighbor problems and performance degradation occur due to heterogeneous workloads consuming shared resources

Engineering Contradiction:

Improveresource utilizationVSAvoidworkload performance

Core Design Contradiction:

Quantity of substanceVSReliability

Solution Approach 1:

The system performs preliminary actions by collecting performance information from hardware performance measurement components and calculating short-term and long-term averages before determining noisy neighbor conditions. This allows the system to establish baseline performance expectations and detect deviations proactively, enabling preventive mitigation before severe resource starvation occurs.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system implements feedback mechanisms by continuously monitoring performance information, comparing it against expected utilization patterns, and using this feedback to identify noisy neighbors. The feedback loop enables dynamic adjustment of resource allocation and mitigation strategies based on real-time performance deviations, resolving the contradiction between resource sharing and performance reliability.

Inventive Principle:

Principle #23Feedback

2Adaptability or versatility

If heterogeneous workloads are allowed to run on shared hardware resources, then workload flexibility and adaptability are improved, but resource starvation and performance issues arise among virtual machines

Engineering Contradiction:

Improveworkload flexibilityVSAvoidvirtual machine performance

Core Design Contradiction:

Adaptability or versatilityVSProductivity

Solution Approach 1:

The system applies dynamics by using moving averages (short-term and long-term) that adapt to changing workload patterns over time. This dynamic approach allows the system to maintain workload flexibility while detecting performance degradation trends, enabling the system to respond appropriately to both legitimate workload variations and noisy neighbor conditions.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system replaces direct mechanical resource allocation controls with a statistical detection approach using performance information collection and analysis. By substituting intrusive control mechanisms with non-intrusive statistical monitoring and comparison against expected patterns, the system maintains workload flexibility while identifying performance issues through data-driven methods.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Measurement precision

If intrusive monitoring and control mechanisms are implemented to detect and mitigate noisy neighbors, then detection accuracy and mitigation effectiveness are improved, but system overhead and complexity increase

Engineering Contradiction:

Improvenoisy neighbor detection accuracyVSAvoidmonitoring system complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system applies self-service by using the hardware performance measurement components to collect performance information that the system itself needs for detection and mitigation. The hardware components perform the measurement functions that would otherwise require complex software monitoring infrastructure, reducing overall system complexity while maintaining detection accuracy.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The system uses parameter changes by transforming raw performance information into normalized metrics through short-term and long-term averaging operations. This parameter transformation simplifies the detection logic by converting complex raw data into comparable statistical measures, reducing the complexity of the monitoring system while improving measurement precision for noisy neighbor detection.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20260104910A1Apparatus, Device, Method, Computer Program and Computer System for Determining Presence of a Noisy Neighbor Virtual Machine

Publication Date: 2026.04.16 INTEL CORP

AI summary

Examples relate to an apparatus, a device, a method, a computer program (or computer-readable medium) and computer system for determining presence of a noisy neighbor virtual machine. Some aspects of the present disclosure relate to an apparatus for a computer system, the apparatus comprising interface circuitry, machine-readable instructions, and processor circuitry to execute the machine-readable instructions to obtain performance information of one or more hardware performance measurement components of the computer system, determine, based on the performance information, a deviation of a utilization of the computer system from an expected utilization of the computer system, and determine presence of a first virtual machine having a workload that impacts a performance of one or more second virtual machines based on the deviation.