Policy-Based Power Management Runtime
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing power management schemes in machines lack the infrastructure to make determinations about power usage based on complex factors, leading to inefficient energy consumption, as they tend to reward early requests over more meritorious later requests, and do not consider the relative value of programs in deciding whether to power devices on or off.
Innovation Solution
Implementing a power management runtime that enforces rich policies to determine whether to change a device's power state, taking into account factors such as program status, current power consumption, energy stored, and future predictions, with a request deflector intercepting requests to prevent unnecessary power usage based on these policies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If traditional power management rules are used to turn devices on or off based on simple inactivity thresholds, then power consumption is reduced through automatic device shutdown, but the system cannot evaluate whether programs merit power usage and tends to reward early requests over more meritorious later requests
Solution Approach 1:
A policy evaluation module is introduced as an intermediary between device requests and power management decisions. This module evaluates whether programs merit power usage by assessing program status, request importance, and system state before allowing devices to power on, thereby enabling sophisticated energy management without sacrificing adaptability
Solution Approach 2:
The system changes the parameter of power state transitions by introducing conditional logic that evaluates multiple factors (program status, request merit, system state) rather than relying on simple inactivity thresholds. This allows the system to make nuanced decisions about when to transition devices between power states based on the relative value of different requests
2Ease of operation
If devices are powered on immediately when any request is made, then system responsiveness and functionality are maintained, but energy is wasted on low-priority requests and early requests are rewarded over more important later requests
Solution Approach 1:
The system changes the parameter of request handling by introducing a merit evaluation mechanism that assesses request importance and program status before triggering device activation. This allows the system to maintain responsiveness for high-priority requests while deferring or denying low-priority requests, thereby reducing energy waste without significantly impacting overall system functionality
Solution Approach 2:
The policy evaluation module autonomously determines whether to allow device activation based on pre-defined criteria for program merit and request importance. This self-service mechanism eliminates the need for immediate device activation for all requests, allowing the system to selectively power devices only when justified by the merit of the request
3Ease of manufacture
If simple power management rules are implemented, then the system is easy to implement and maintain, but it lacks the infrastructure to make determinations about power usage based on rich or complex factors
Solution Approach 1:
The power management system is segmented into distinct modular components: a policy evaluation module that assesses request merit, a decision module that determines power state transitions, and device control mechanisms. This segmentation allows the system to implement sophisticated evaluation logic while maintaining ease of implementation through modular, independently configurable components
Data Source
AI summary
A machine's power usage may be managed by a power-management policy. When a program makes a request that involves use of one of the machine's power-consuming devices, the policy may take into account factors such as the program's status, where the status indicates the program's relative level of justification to consume power. A component may intercept a request to use a device before the request reaches the driver, and may deflect requests that, if carried out, are not consistent with power usage policy. Infrastructure supports the use of policies that determine whether a particular device's power state will be changed.