Data Center Power Control via Dynamic Task Migration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data centers face inefficiencies in managing energy consumption and costs due to varying customer quality of service requirements, leading to increased power usage and operational burdens, without considering the true cost of operations or energy consumption.
Innovation Solution
A power control system that utilizes virtualization and a service/power controller to optimize processing tasks based on power consumption and performance requirements, balancing quality of service demands with energy management, and adjusting operations to minimize power usage and costs by monitoring and predicting power consumption and external factors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data centers meet quality of service requirements by running servers at high load, then customer service levels are improved, but energy consumption increases
Solution Approach 1:
The patent implements dynamic task distribution that adjusts server loading based on predicted energy costs and quality of service requirements. The system dynamically migrates tasks between servers and adjusts processing priorities to meet customer SLAs while minimizing energy consumption during high-cost periods.
Solution Approach 2:
The system changes operational parameters by adjusting task distribution, server consolidation, and processing priorities based on energy cost predictions. It modifies system behavior in response to varying energy prices, time of day, and customer service requirements to optimize the balance between quality of service and energy consumption.
2Reliability
If data centers run servers constantly to meet steady state quality of service requirements, then service availability is maintained, but operational costs increase
Solution Approach 1:
The system performs preliminary actions by predicting future energy costs and proactively adjusting task distribution before high-cost periods occur. It anticipates energy price variations and pre-positions tasks on energy-efficient servers or consolidates workloads to avoid high operational costs while maintaining service availability.
Solution Approach 2:
The system implements feedback mechanisms that continuously monitor actual energy consumption, compare it with predicted costs, and adjust task distribution accordingly. This closed-loop control ensures that service level agreements are met while operational costs are optimized based on real-time performance data.
3Productivity
If data centers allocate more processing power to meet customer demands, then quality of service improves, but burden on power resources increases
Solution Approach 1:
The patent segments the data center into multiple servers with varying power characteristics and efficiency profiles. By dividing the processing workload across these segmented units and selectively activating only the most efficient servers for specific task types, the system maintains high productivity while reducing overall power consumption.
Solution Approach 2:
The system creates a universal task distribution framework that can adapt to different customer requirements, task types, and energy cost conditions. The same infrastructure serves multiple customers with different quality of service needs while dynamically optimizing power usage based on current operational conditions and energy prices.
4Ease of operation
If data centers monitor and optimize based on true operational costs, then decision-making improves, but system complexity increases
Solution Approach 1:
The patent introduces an intermediary layer (the service/power controller) that handles the complexity of cost calculation, prediction, and optimization. This intermediary translates complex energy cost data and system state information into simple task distribution decisions, improving decision-making quality while containing system complexity within a dedicated management component.
Data Source
AI summary
A power control system in a data center has a plurality of physical servers, each server having a local controller, at least one virtual server coupled to at least some of the physical servers, and a central controller to control task loading on the physical servers through the virtual servers. A method of controlling power consumption in a data center includes receiving inputs from local controllers residing on loads, the inputs including data about power consumption on the loads, receiving as an input at least one quality of service requirement, and allocating tasks to at least one server based upon the quality of service and the power consumption on the loads.