Workload Power Attribution via Load-Based Apportionment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional power monitoring systems do not provide effective methods for attributing electrical power consumption to workloads such as virtual machines and virtualized entities in computing environments, making it challenging to manage and optimize power usage efficiently.
Innovation Solution
A method and system that utilize a power apportionment module to determine and attribute a portion of a server's electrical power consumption to specific workloads, such as software applications, web applications, and virtual machines, based on load levels and usage metrics, using sensors to measure power consumption and apply power consumption contribution models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional power meters or sensors are used to monitor electrical power consumption of hardware, then power levels and computational loading can be monitored, but there is no effective method for attributing power consumption to workloads such as virtual machines and virtualized entities
Solution Approach 1:
The patent segments the server's total power consumption into distinct portions attributable to different workloads by using power consumption contribution models. The system divides the overall power measurement into workload-specific allocations based on CPU usage, memory usage, and other performance metrics, enabling precise workload-level power accounting without requiring separate physical sensors for each virtual entity.
2Productivity
If power monitoring systems are implemented at the hardware level, then power usage can be managed and optimized, but analogous systems for monitoring workload power usage are not provided
Solution Approach 1:
The patent introduces power consumption contribution models as intermediary computational structures that bridge hardware power measurements and workload-level accounting. These models act as mediators that translate physical power consumption data into workload-attributed values, enabling virtualized workload monitoring without requiring complex hardware modifications or separate sensing infrastructure for each virtual entity.
3Use of energy by moving object
If workload power consumption is not attributed to specific workloads, then resource allocation and cost tracking cannot be optimized, but implementing attribution systems increases system complexity
Solution Approach 1:
The patent changes the parameters used for power consumption accounting by introducing performance-based metrics (CPU usage percentage, memory usage, disk I/O, network bandwidth) as weighting factors in power consumption contribution models. This allows the system to dynamically adjust power attribution based on actual workload performance parameters, enabling energy optimization while keeping the system relatively simple through software-based calculation rather than complex hardware modifications.
Data Source
AI summary
A method for attributing a portion of a level of electrical power consumption by a server to a virtual machine executing on the server, includes determining a level of electrical power consumption of a server. The method includes identifying a level of load placed on the server by a workload executing on the server. The method includes attributing, by a power apportionment module in communication with the server, a portion of the level of electrical power consumption of the server to the workload, based on the identified level of load. The method includes providing, by the power apportionment module, to a user, an identification of the attributed portion of the level of electrical power consumption.