Hierarchical Power Budgeting for Server Density and Reliability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data centers face challenges in maximizing server density due to power consumption and cooling limitations, leading to increased costs and potential service outages, as existing methods like de-rating and estimating power consumption based on name plate ratings can be inaccurate and fail to adapt to changing workloads.
Innovation Solution
Implementing a hierarchical power budgeting system that allocates and dynamically adjusts power consumption across zones, racks, and servers using management processors, allowing for active management of power consumption based on actual needs and workload changes, utilizing power-regulation states and clock modulation to optimize power usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If server density is increased by filling blade server chassis with multiple server modules, then computing capacity per unit space is improved, but cooling costs increase and service outages become more frequent
Solution Approach 1:
The patent implements dynamic power management by allowing servers to transition between different power states (P-states) based on actual workload demands. The system dynamically adjusts CPU frequency and voltage levels in response to monitored workload changes, enabling the server to operate at optimal power consumption levels rather than maintaining fixed high power consumption, thus reducing cooling requirements and improving reliability while maintaining high computing capacity.
Solution Approach 2:
The patent changes key operating parameters including CPU frequency, voltage, and power state assignments to optimize the balance between computing performance and power consumption. By implementing variable frequency voltage scaling and dynamic power state transitions, the system adapts parameters in real-time to match actual workload requirements, reducing overall power consumption and cooling demands while maintaining high server density.
2Device complexity
If name plate power ratings are used to estimate power consumption, then power allocation is simplified, but actual power consumption is mis-predicted leading to inefficient resource allocation
Solution Approach 1:
The patent implements a feedback mechanism where the system continuously monitors actual power consumption and compares it against allocated power budgets. Based on this feedback, the system dynamically adjusts power state assignments and workload distribution to optimize power utilization. This closed-loop approach enables accurate real-time power consumption tracking and adaptive resource allocation, eliminating the inaccuracies of static name plate ratings while maintaining manageable system complexity through automated management.
3Device complexity
If fixed power allocation is used for servers, then power management is simplified, but unexpected workload changes cause power demand to exceed allocated power leading to circuit breaker trips
Solution Approach 1:
The patent replaces fixed power allocation with dynamic power management that continuously adapts to changing workload conditions. The system monitors actual power consumption and dynamically adjusts power state assignments, CPU frequency, and voltage levels in real-time. This dynamic approach ensures that power allocation automatically responds to unexpected workload changes, preventing circuit breaker trips and maintaining power supply stability without requiring complex manual intervention.
Solution Approach 2:
The patent implements preliminary power state configuration and reservation mechanisms where power budgets are pre-established but dynamically adjustable. The system maintains reserved power capacity and can preemptively adjust power allocations before actual outages occur by monitoring trends and anticipating workload changes. This proactive approach ensures power supply stability by preparing the power management system in advance for potential demands while keeping management complexity manageable through automation.
Data Source
AI summary
A method for adjusting power consumption during server operation is disclosed. The method comprises initializing the server to a system power performance state (SPP-state) that comprises a plurality of constituent power settings, determining whether power regulation is required while the server is operating at the current SPP-state, setting a flag if power regulation is required, and increasing at least one of the plurality of constituent power settings associated with the SPP-state if the flag has been set.