Multi-Server Load Balancing Using Power Supply Efficiency Curves
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for reducing server power consumption in data centers are inefficient and do not effectively minimize overall power consumption or heat generation, despite efforts to improve cooling efficiency and equipment replacement.
Innovation Solution
A method and system for intelligently balancing workloads across multiple servers based on the efficiency curves of power supply units, using a manager to monitor and dynamically adjust load distribution to optimize power consumption and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Temperature
If servers are placed in different locations to improve cooling airflow path, then cooling efficiency is improved, but power consumption is not significantly reduced
Solution Approach 1:
The patent changes the operational parameters of servers by adjusting workload distribution based on real-time power consumption characteristics and efficiency curves. The manager dynamically modifies load parameters across servers to optimize overall power consumption rather than relying on physical repositioning for cooling efficiency.
2Use of energy by moving object
If old equipment is replaced with new efficient equipment, then power consumption is reduced, but system complexity and cost increase
Solution Approach 1:
The manager software provides multi-functionality by performing workload distribution, power consumption monitoring, efficiency curve generation, and dynamic load balancing across heterogeneous servers. This universal management system optimizes power consumption without requiring replacement of physical equipment, reducing system complexity while maintaining energy efficiency.
3Use of energy by moving object
If underutilized servers are eliminated, then power consumption is reduced, but system reliability decreases
Solution Approach 1:
The system dynamically adjusts workload distribution based on real-time power consumption characteristics and efficiency curves. Servers that appear underutilized can be dynamically activated when their efficiency characteristics indicate they would be optimal for specific workloads, maintaining system reliability while optimizing power consumption through dynamic rather than static server management.
4Speed
If load balancing is performed without considering power consumption characteristics, then system responsiveness is maintained, but overall power consumption is not optimized
Solution Approach 1:
The manager implements feedback mechanisms by continuously monitoring real-time power consumption characteristics of servers, generating efficiency curves, and using this feedback to dynamically adjust load balancing decisions. This closed-loop system optimizes overall power consumption while maintaining system responsiveness by considering both performance and power characteristics in load distribution.
Data Source
AI summary
A method for reducing a total power consumption of multiple servers includes determining an efficiency distribution of a power supply unit of each of the multiple servers, the efficiency distribution including a working efficiency of the power supply unit varied according to a utilization ratio of the power supply unit, wherein the efficiency distribution of the power supply unit of at least one of the multiple servers is obtained based on a measurement performed by a BMC; retrieving, by a manager, the efficiency distribution of the power supply unit of each of the multiple servers; and performing load balancing on the multiple servers through the manager to reduce the total power consumption of the multiple servers. The load balancing is performed based at least in part on the efficiency distribution of the power supply unit of one or more of the multiple servers.