Cloud Storage Scheduling with Predictive IOPS Limits
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Cloud computing platforms experience reduced storage efficiency due to overloaded virtual disks in instances, affecting the performance of other instances and the overall system.
Innovation Solution
Implement a scheduling module that uses a predictive IOPS value model to set a maximum IOPS value for virtual disks, limiting them when overloaded, and removing the limitation when normal operation is restored, thereby optimizing storage resource allocation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the cloud computing platform sets quotas to each instance for scheduling storage resources, then storage resource allocation is managed, but when the virtual disk in an instance is overloaded, the performances of other instances are reduced and storing efficiency is reduced
Solution Approach 1:
The scheduling module continuously monitors IOPS values of virtual disks and dynamically adjusts quotas based on real-time feedback. When a virtual disk's IOPS exceeds a threshold, the system automatically reduces its quota to prevent overload, and restores the quota when the disk returns to normal operation, creating a closed-loop control system that balances storage efficiency and performance stability
Solution Approach 2:
The patent implements dynamic quota adjustment where the IOPS quota for each instance is not fixed but changes based on real-time monitoring of virtual disk performance. The system transitions from static quota allocation to dynamic quota management, allowing the cloud platform to adapt to changing workload conditions and prevent overload-induced performance degradation
Data Source
AI summary
A method of optimizing the scheduling of storage resource for a server used in cloud computing obtains a first input/output operations per seconds (IOPS) value of a virtual disk in an instance. A warning message is generated when the first IOPS value is larger than a predefined threshold value, which represents an overloaded state of the virtual disk. An predicted IOPS value is obtained and the predicted IOPS value is set as the maximum IOPS value for limiting the IOPS value of the virtual disk, thus other virtual disks are not affected by a virtual disk being overloaded, and storing efficiency of the server is improved. A server and a computer readable storage medium applying the method are also disclosed.