Dynamic Power Supply for Server Farms
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data centers face significant challenges in managing power consumption and operational costs due to over-provisioning and inefficient power usage, with existing solutions only addressing partial aspects of power management for individual servers rather than entire grids or systems.
Innovation Solution
A system comprising a power supply with multiple generators and a power manager that dynamically adjusts power output based on system requirements, coupled with a global power manager that redistributes workload across computers to minimize power consumption, allowing for flexible power allocation and reduction in operational costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If servers are over-provisioned to meet peak demand, then reliability is improved, but power consumption increases
Solution Approach 1:
The patent implements dynamic power management by enabling servers to transition between different power states (active, idle, hibernation) based on real-time workload demands. The system continuously monitors workload levels and automatically adjusts server power consumption, allowing servers to operate at full power only when necessary to meet peak demand, while consuming minimal power during low-utilization periods.
Solution Approach 2:
The patent creates a universal power management system that can handle multiple server states and workload types through a single centralized controller. This global power manager coordinates power allocation across the entire server farm, enabling any server to serve any workload type, thereby optimizing overall power utilization while maintaining the ability to meet peak demand across diverse computational requirements.
2Productivity
If servers are designed to run at maximum power levels, then productivity is improved, but use of energy worsens
Solution Approach 1:
The system dynamically adjusts server power consumption based on actual workload requirements rather than operating at fixed maximum levels. The global power manager continuously monitors productivity metrics and power consumption, adjusting server operational states to maintain required computing performance while minimizing energy usage during periods of lower demand.
Solution Approach 2:
The patent changes the operational parameters of servers by introducing multiple power states (full power, partial power, idle, hibernation) instead of a single maximum power mode. This allows the system to optimize the balance between productivity and power consumption by selecting appropriate power levels based on real-time workload characteristics and performance requirements.
3Use of energy by moving object
If individual server power management is implemented, then use of energy is improved, but device complexity increases
Solution Approach 1:
The patent merges individual server power management functions into a single centralized global power manager. Instead of each server having independent complex power management systems, the centralized controller consolidates power management intelligence, reducing overall system complexity while maintaining or improving power efficiency through coordinated control of the entire server farm.
Solution Approach 2:
The global power manager acts as an intermediary between the workload demands and individual servers, simplifying the power management architecture. Rather than implementing complex power management logic in each server, the intermediary controller handles all power state transitions and workload allocations, reducing device complexity at the server level while maintaining centralized optimization capability.
4Productivity
If more servers are added to meet growing demand, then productivity is improved, but use of energy worsens
Solution Approach 1:
The patent merges multiple server workloads onto fewer servers through virtualization and workload consolidation techniques. By combining computational tasks from multiple physical servers into optimized configurations on fewer machines, the system maintains or improves total computing capacity while reducing the number of active servers and their associated power consumption.
Solution Approach 2:
The system creates universal servers capable of handling multiple different workload types through virtualization. A single physical server can dynamically serve multiple virtual machines with different computational requirements, thereby maintaining high productivity across diverse workloads while minimizing the total number of physical servers needed and their collective power consumption.
Data Source
AI summary
A computer is disclosed with an operating system including a kernel and a task scheduler to schedule execution of one or more processes on the computer; a power estimator is coupled to the task scheduler to determine a required system power based on the number or type of processes scheduled for execution; and a variable load power supply including a plurality of power generators each having a predetermined power output and a power manager receiving instructions from the power estimator in the kernel, the power manager enabling a sufficient number of power generators to match the required power generated by the power estimator.