Storage System Power Consumption Control via Compute Node Shutdown
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data centers face challenges in controlling power consumption, particularly from storage systems, which are significant consumers of electric power, leading to potential power peaks. There is a need for an effective method to manage and limit power usage within these systems.
Innovation Solution
A method and system for power consumption control in storage systems, which involves monitoring power usage across racks containing compute nodes and storage nodes. When power thresholds are exceeded, measures such as shutting down compute nodes or reducing write operations to storage nodes are implemented to manage power consumption, with a power manager module determining the most effective actions based on current usage patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If storage systems operate at full capacity to maintain system performance, then productivity is improved, but power consumption increases leading to power peaks
Solution Approach 1:
The patent implements dynamic power management by continuously monitoring power consumption levels and adjusting compute node operations and write bandwidth in real-time. The system transitions between different operational states based on current power conditions, enabling adaptive control that balances performance and power usage rather than operating at fixed capacity levels
Solution Approach 2:
The system changes operational parameters such as write bandwidth limits and compute node activity levels in response to power consumption thresholds. By dynamically adjusting these parameters, the system can reduce power consumption during peak periods while maintaining adequate performance during normal operation, directly addressing the contradiction between sustained high performance and power consumption control
2Use of energy by moving object
If power consumption is reduced by shutting down compute nodes or reducing write operations, then power consumption is controlled, but system performance deteriorates
Solution Approach 1:
The patent employs feedback mechanisms where power consumption is continuously monitored and this information feeds back into control decisions. The system adjusts compute node operations and write bandwidth based on real-time power measurements, creating a closed-loop control system that automatically balances power consumption with performance requirements rather than using fixed reduction strategies
Solution Approach 2:
Instead of completely shutting down compute nodes or stopping write operations, the system applies partial actions by reducing write bandwidth to specific storage nodes or throttling compute node activity partially. This approach maintains sufficient system performance while achieving power consumption targets, avoiding the extreme of complete shutdown that would severely impact productivity
3Use of energy by moving object
If dynamic adjustment of compute node operations is implemented to reduce power consumption, then power consumption is controlled, but device complexity increases
Solution Approach 1:
The patent divides the storage system into manageable segments including compute nodes, storage nodes, and a power manager module. Each segment can be independently monitored and controlled, allowing complex power management tasks to be broken down into simpler, modular operations that are easier to implement and manage than a monolithic control system
Data Source
AI summary
A method for power management of a storage system unit, the method may include selecting a power reduction measure out of (a) a shutdown of at least one stateless compute node out of stateless compute nodes of the storage system unit, and (b) an other power reduction measure that does not involve the shutdown of the at least one stateless compute node, while prioritizing the shutdown of the at least one stateless compute node over the other power reduction measure; wherein the storage system unit further comprises storage nodes; wherein states related to communications with client computers (i) are maintained in the storage nodes, and (ii) are not stored on the stateless compute nodes; and applying the power reduction measure based on the selecting.