Cluster Power Control via Battery Switching and DVFS
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional power management technologies for data centers face challenges in reducing power consumption and actively managing availability in cluster systems, as they often require constant server power and are inefficient in handling power fluctuations.
Innovation Solution
A power control apparatus and method that monitors real-time power consumption and usage in cluster systems, using Dynamic Voltage and Frequency Scaling (DVFS) and a battery control system to set power capping thresholds, automatically adjusting power supply by turning on/off batteries and changing server states to maintain optimal power usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If conventional power management technologies are used to control power in data centers, then power consumption can be managed to some extent, but the system cannot actively cope with availability and requires servers to always remain powered on
Solution Approach 1:
The system performs preliminary actions by monitoring power consumption trends and server usage patterns to predict when power capping thresholds will be reached. This allows the system to prepare battery backup units in advance and smoothly transition servers to hibernation states before power shortages occur, maintaining availability while managing power consumption effectively
Solution Approach 2:
The system dynamically adjusts power management strategies based on real-time monitoring of power consumption and server usage. It continuously adapts by transitioning servers between active and hibernation states, adjusting battery discharge rates, and modifying power allocation based on current system conditions, thereby balancing power consumption reduction with availability maintenance
2Loss of energy
If servers are powered down or converted to hibernation state to reduce power consumption, then power usage decreases, but the complexity required to return servers to original state increases and efficiency decreases
Solution Approach 1:
The system implements self-service mechanisms where servers automatically monitor their own power consumption and usage patterns, making autonomous decisions about when to enter hibernation or shut down based on predefined criteria. This reduces the complexity of manual server management and enables automated restoration processes that minimize intervention requirements
Solution Approach 2:
The system employs continuous feedback loops that monitor server state, power consumption, and system availability. When servers are powered down or hibernated, the feedback mechanism tracks their status and automatically triggers restoration processes when needed, reducing the complexity of manual state management and ensuring efficient server reuse
3Loss of energy
If power capping threshold is set low to reduce power consumption, then energy usage decreases, but system availability and performance may be compromised
Solution Approach 1:
The system applies local quality by implementing differentiated power management strategies for different servers and workloads within the data center. Instead of uniformly applying power caps, it identifies and prioritizes critical servers that maintain system functionality while allowing non-critical servers to be powered down or hibernated, thereby reducing overall power consumption without compromising essential system performance
Solution Approach 2:
The system dynamically changes operational parameters such as power allocation, server states, and battery discharge rates based on real-time conditions. It adjusts the power capping threshold and related parameters according to system availability requirements and workload characteristics, enabling flexible optimization of power consumption while maintaining adequate system performance
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach reduces power consumption, minimizes the occurrence of power-related issues, and dynamically manages power based on real-time monitoring, enhancing the availability and efficiency of cluster systems.
Implementation Method 1
a battery mounted in a power supply unit of the cluster system
Implementation Method 2
changing frequency using Dynamic Voltage and Frequency Scaling (DVFS)
Data Source
AI summary
A power control apparatus for a cluster system, includes a cluster including a plurality of nodes, each equipped with a battery; and a power control unit connected to the cluster over a network and configured to monitor power management information and performance information of the cluster and to set a power capping threshold based on the monitored power management information and performance information of the cluster. Accordingly, the power control unit enables power of the cluster to be limited by turning on and off the batteries when power of the cluster system increases up to the power capping threshold.