Cluster Power Management via Dynamic Reconfiguration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Computer clusters consume excessive power due to being designed for worst-case data processing loads, leading to inefficiencies as not all computers operate at optimal levels, and there is no method to account for varying power efficiencies of different computers in cluster configuration.
Innovation Solution
Implement a power management system that reconfigures the cluster by selecting a configuration of computers capable of supporting a processing load at the lowest power level, using a system management node to compare and manage power consumption across different configurations, and dynamically adjust the cluster's configuration based on changing processing loads.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If computers in a cluster are designed for worst-case data processing loads, then the cluster can handle peak processing demands, but power consumption increases significantly as not all computers operate at optimal levels
Solution Approach 1:
The patent implements dynamic cluster configuration where the system management node continuously monitors processing loads and dynamically reconfigures the cluster by adding or removing computers based on actual demand. This dynamic adjustment allows the cluster to optimize power consumption by having fewer computers active during low-demand periods while maintaining the ability to handle peak loads when necessary.
Solution Approach 2:
The system changes the operational parameters of the cluster by adjusting the number of active computers, their processing capacities, and power consumption levels based on real-time processing load requirements. The system management node evaluates different configuration scenarios and selects optimal parameters that balance processing capacity needs with power consumption efficiency.
2Adaptability or versatility
If computers are selected for cluster inclusion based on general availability, then more resources can be pooled, but the varying power efficiencies of different computers cannot be accounted for in configuration
Solution Approach 1:
The system management node receives feedback about the processing loads actually being executed by the cluster and uses this information to evaluate different computer configurations. This feedback mechanism allows the system to account for the varying power efficiencies of different computers by monitoring actual performance and adjusting configurations to maximize power efficiency while maintaining adequate processing capacity.
Solution Approach 2:
The system performs preliminary evaluation of different cluster configurations by considering the power efficiency characteristics of available computers before finalizing the cluster composition. The system management node proactively selects configurations that optimize power efficiency based on predicted processing loads and the known efficiency characteristics of available computer resources.
3Device complexity
If a fixed cluster configuration is used, then system simplicity is maintained, but the cluster cannot optimize power consumption for changing processing loads
Solution Approach 1:
The system management node autonomously monitors processing loads, evaluates configuration options, and reconfigures the cluster without requiring manual intervention. This self-service capability allows the system to automatically optimize power consumption by detecting when processing loads change and initiating appropriate reconfiguration actions, thereby reducing the complexity of manual configuration management while achieving dynamic power optimization.
Data Source
AI summary
Power management for clusters of computers, a cluster including a configuration of computers operating in a power management domain, one or more of the computers of the cluster having a power requirement that differs from power requirements of other computers in the cluster, data processing operations on the cluster effecting a processing load on the computers in the cluster, including selecting, from a plurality of configurations of computers capable of supporting a plurality of processing loads at a plurality of power levels, a configuration of computers capable of supporting the processing load at a lowest power level, the lowest power level defined by comparison among configurations capable of supporting the processing load; and reconfiguring the cluster according to the selected configuration.