Virtual Machine Power Management via Shared Memory Coordination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In virtualized computing environments, especially on mobile platforms, power management (PM) becomes complex due to shared system resources among virtual machines (VMs), leading to inefficiencies as local PM decisions by VMs are not accurately translated into hardware power savings, causing high system overheads and performance limitations.
Innovation Solution
A distributed and secure power management architecture that uses shared memory objects to allow VMs to coordinate PM decisions, enabling them to access and share PM-related information, thus making informed decisions without VMM intervention, resulting in direct hardware PM requests.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If VMs make local PM decisions independently, then PM decisions can be made quickly, but power savings are not accurately achieved due to shared resources
Solution Approach 1:
The patent merges local PM decisions with global resource awareness by implementing a shared memory structure that stores PM state information for all vCPUs. This allows each VM to access the complete picture of resource usage across all VMs sharing the same physical CPU, enabling coordinated PM decisions that balance speed and energy efficiency.
Solution Approach 2:
The shared memory structure acts as an intermediary between individual VMs and the physical hardware. It provides each VM with visibility into the PM state of other vCPUs without requiring complex inter-VM communication protocols, thus achieving coordinated power management while maintaining decision speed.
2Loss of energy
If VMM mediates all PM decisions, then globally-optimal PM decisions can be made, but system overhead increases due to hypercalls
Solution Approach 1:
The patent enables self-service by providing each VM with direct access to shared memory containing PM state information. VMs can autonomously make informed PM decisions based on global resource usage patterns without needing to query the VMM, thereby reducing hypercall overhead while maintaining optimal power management.
Solution Approach 2:
The shared memory structure provides continuous feedback to each VM about the PM state of other vCPUs. This feedback mechanism enables VMs to adjust their power management decisions based on real-time resource usage patterns across all VMs, achieving global optimality without VMM intervention.
3Loss of energy
If frequent PM requests are made to maximize battery life, then power management accuracy improves, but performance and scalability are limited due to overhead
Solution Approach 1:
The patent implements preliminary action by pre-establishing a shared memory structure that continuously stores PM state information for all vCPUs. This pre-computed information is immediately available to each VM when making PM decisions, eliminating the need for frequent queries to the VMM and reducing system overhead while maintaining accurate power management.
Solution Approach 2:
The patent replaces the mechanical system of frequent hypercalls to the VMM with a more efficient information sharing mechanism using shared memory. This substitution allows VMs to access PM state information directly without the overhead of context switches and hypercall processing, thus improving performance while maintaining battery life optimization.
Data Source
AI summary
A method and apparatus for cooperative power management by virtual machines is disclosed. In an example apparatus, a computing device includes a number of virtual machines (VMs), each sharing a resource with another of the VMs. A virtual machine monitor (VMM) includes a shared memory object accessible by the VMs, wherein the shared memory object comprises a current power state requested by a VM for the resource.