Oversubscribed VM Resource Scheduling via Proactive Migration

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

In cloud computing, oversubscription of virtual machines leads to unpredictability in resource availability, causing degradation in user experience due to unexpected resource usage spikes, as resources are often limited or throttled to accommodate other virtual machines on the same physical host.

Innovation Solution

A usage spike alerting service that allows users to request additional computing resources for a specified future time window, enabling proactive migration of virtual machines to less crowded hosts, thereby ensuring resource availability during predicted spikes without interrupting the workload.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If virtual machines are oversubscribed on the same physical host to improve resource utilization, then resource efficiency is improved, but resource availability becomes unpredictable and user experience degrades

Engineering Contradiction:
Improveresource utilizationVSAvoidresource availability
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The system performs proactive migration of virtual machines to different physical hosts before resource conflicts occur. By monitoring resource usage patterns and predicting future conflicts, the system preemptively relocates virtual machines to ensure resource availability during high-demand periods, thus maintaining both high utilization and reliable availability.

Inventive Principle:
Principle #10Preliminary action

2Reliability

If additional computing resources are allocated to virtual machines to ensure availability during high-demand periods, then resource availability is improved, but overall resource utilization efficiency decreases

Engineering Contradiction:
Improveresource availabilityVSAvoidresource utilization
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The system dynamically adjusts the allocation and location of virtual machine resources based on real-time and historical usage patterns. Instead of static over-provisioning, the system continuously monitors resource demand and migrates virtual machines between hosts to match actual usage patterns, ensuring availability when needed while maintaining high overall utilization efficiency.

Inventive Principle:
Principle #15Dynamics

3Reliability

If virtual machines are migrated proactively to less crowded hosts to ensure resource availability, then user experience is improved, but system complexity increases

Engineering Contradiction:
Improveuser experienceVSAvoidsystem complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The system implements automated self-service mechanisms for virtual machine migration. The resource management system autonomously monitors usage patterns, predicts conflicts, selects appropriate target hosts, and executes migrations without manual intervention. This automation handles the complexity internally while presenting a simple, reliable service to users.

Inventive Principle:
Principle #25Self-service

Data Source

PatentUS11249790B1Scheduling usage of oversubscribed computing resources
Publication Date: 2022.02.15 AMAZON TECH INC
  • US11249790B1 patent drawing
  • US11249790B1 patent drawing
  • US11249790B1 patent drawing

AI summary

Generally described, one or more aspects of the present application relate to an instance resource oversubscription service for scheduling a burst period for a running virtual machine instance based on a time window specified by a user of the virtual machine instance. For example, the instance resource oversubscription service can predict future resource usage and identify the appropriate timing and physical host machine for letting the user burst (e.g., temporarily use the virtual machine instance in a manner than consumes a higher amount of computing resources such as CPU cycles, memory, network bandwidth, etc.). In doing so, the instance resource oversubscription service may consider, for example, the historical and current resource utilization levels of the virtual machine instances running on a set of available physical host machines and the burst period scheduling requests from other users of the instance resource oversubscription service.