Dynamic Disk Grouping for Energy-Saving Storage Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Storage systems in data centers face high energy consumption and carbon emissions due to continuous IO loads, where existing energy-saving methods fail to effectively reduce power usage as they trigger disk sleep based on fixed thresholds, affecting service performance and not matching IO loads dynamically.
Innovation Solution
A storage system is designed with multiple disk groups, where one group processes IO services, another is in an energy-saving state, and a third is in a sleep state, with dynamic grouping based on predicted IO loads and access frequencies to optimize energy usage and resource allocation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If disk sleep is triggered based on fixed idle duration threshold, then energy saving is attempted, but service performance deteriorates and energy-saving effect is reduced due to continuous IO loads
Solution Approach 1:
The patent implements dynamic disk grouping based on predicted IO loads rather than using fixed thresholds. The system continuously predicts future IO loads and dynamically adjusts disk group configurations, transitioning disks between active and sleep states based on predicted rather than static criteria, thereby adapting to changing workload patterns and achieving both energy savings and service performance maintenance
Solution Approach 2:
The system performs preliminary prediction of IO loads before making disk sleep/wake decisions. By predicting future IO patterns in advance, the system can proactively group disks and adjust their states before actual IO requests occur, preventing premature wake-ups and ensuring energy-saving opportunities are captured without compromising service performance
2Device complexity
If all disks are divided into RAID groups and sleep is triggered only when all disks are idle, then disk management is simplified, but energy-saving effect is poor due to continuous IO loads in at least one disk
Solution Approach 1:
The patent segments disks into multiple independent groups (first disk group, second disk group, third disk group) with different functional roles. This segmentation allows independent control of each group's sleep/wake state based on their specific IO load patterns, enabling finer-grained energy management compared to treating all disks as a single RAID group, thereby reducing overall energy consumption while maintaining manageable complexity through automated grouping
3Loss of energy
If disks are dynamically grouped based on predicted IO loads, then energy-saving control precision is improved, but system complexity increases
Solution Approach 1:
The system implements a feedback mechanism where predicted IO loads continuously inform disk group reconfiguration decisions. The controller monitors actual IO patterns, compares them with predictions, and uses this feedback to refine future predictions and adjustment decisions, creating a self-learning system that improves accuracy over time while managing complexity through automated control loops
Data Source
AI summary
This application provides a storage system includes disks grouped into disk groups. One disk group processes an IO service, and another disk group is in an energy-saving state or a sleep state and provides a standby disk for a previous disk group. IO load is focused on a disk in one disk group, so that a requirement for processing the IO service is satisfied by using this group of disks. In addition, the another disk group may save energy or sleep, so that more energy consumption can be reduced. Moreover, resource pools of different levels are implemented by using a plurality of disk groups. When a disk in a previous disk group is in shortage, a capacity of the previous disk group can be expanded by using a next disk group.