JVM Power Manager for Dynamic Voltage and Frequency Scaling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In multi-threaded environments, dynamic voltage and frequency scaling (DVFS) can lead to extended execution times for computation-intensive applications due to inconsistent processor frequency adjustments across threads, as one thread's reduction in frequency may not meet the needs of others.
Innovation Solution
A power manager module within a JAVA Virtual Machine (JVM) implements dynamic processor supply voltage and frequency scaling, using performance monitoring counters and a Markov model to predict processor utilization and adjust clock frequency and voltage accordingly, ensuring optimal energy consumption while minimizing execution time impacts across multiple threads.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If the processor supply voltage is reduced to decrease power dissipation, then energy consumption is reduced, but execution time increases due to increased delay through the device
Solution Approach 1:
The system dynamically adjusts processor frequency and voltage based on real-time workload conditions. The power manager continuously monitors processor utilization and transitions between different power states (C-states and P-states) to optimize the balance between power consumption and execution performance, applying dynamics by making the system adaptable rather than static
Solution Approach 2:
The invention changes physical parameters (frequency and voltage) of the processor to resolve the contradiction. By adjusting these parameters based on workload demands, the system can reduce power dissipation when high performance is not needed while maintaining execution time when computation-intensive tasks are detected
2Use of energy by moving object
If one user application scales down the processor supply voltage and clock frequency to reduce power consumption, then energy efficiency improves, but execution time of computation-intensive applications increases
Solution Approach 1:
The power manager implements feedback mechanisms by monitoring processor utilization through performance monitoring counters and using Markov models to predict future workload states. This feedback loop allows the system to make informed decisions about when to scale voltage and frequency, ensuring that power-saving actions do not negatively impact computation-intensive applications
Solution Approach 2:
The system performs preliminary action by using the Markov model to predict future processor utilization states before actually transitioning power states. This allows the system to proactively adjust frequency and voltage in anticipation of workload changes, preventing execution time delays before they occur
3Use of energy by moving object
If DVFS is performed in hardware without software involvement to reduce power dissipation, then power optimization is achieved, but adaptability to multi-threaded workloads decreases
Solution Approach 1:
The invention introduces an intermediary layer (the power manager in JVM) between the hardware DVFS mechanism and the multi-threaded applications. This intermediary monitors workload characteristics across multiple threads and makes intelligent decisions about frequency scaling, combining hardware efficiency with software-level workload awareness to maintain both power optimization and adaptability
Data Source
AI summary
A machine learning technique is used to improve dynamic prediction of processor utilization for multi-threaded user-level applications in a dynamic run-time environment based on processor utilization history. Processor supply voltage and processor clock frequency may be dynamically scaled based on the predicted processor utilization in order to reduce processor power consumption.