Reinforcement Learning Backup Scheduling for IO Resource Conflicts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data backup systems face resource conflicts between data backup tasks and input/output (IO) tasks, leading to potential performance degradation and errors in service tasks due to inadequate scheduling of replication tasks.
Innovation Solution
A reinforcement learning model is used to generate candidate backup policies based on system states, determining reward values to select a target backup policy that avoids resource conflicts and ensures efficient data backup without affecting IO tasks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data backup tasks are performed using traditional scheduling methods, then backup operations can be completed, but resource conflicts occur between backup tasks and IO tasks leading to performance degradation
Solution Approach 1:
The patent implements a reinforcement learning model that continuously monitors system state and adjusts backup policies based on real-time feedback about resource usage and task performance. The model learns from historical data about IO task patterns and dynamically modifies backup scheduling to avoid resource conflicts, thereby maintaining IO performance while ensuring backup completion.
Solution Approach 2:
The backup policy is transformed from a static, pre-defined schedule to a dynamic, adaptive policy generated by the reinforcement learning model. The model continuously generates new backup policies based on current system state, allowing the backup operations to adapt to changing IO task demands and resource availability, thus avoiding performance degradation.
2Adaptability or versatility
If administrators manually arrange replication tasks according to different needs, then different data recovery objectives can be achieved, but the system complexity and scheduling difficulty increase
Solution Approach 1:
The reinforcement learning model autonomously generates and optimizes backup policies without requiring manual administrator intervention. The system self-adjusts to different data recovery objectives by learning from historical performance data and automatically modifying scheduling parameters, thereby reducing operational complexity while maintaining versatility.
Solution Approach 2:
The patent changes the approach from manual parameter configuration to automated parameter optimization. The reinforcement learning model dynamically adjusts scheduling parameters such as backup timing, resource allocation, and priority levels based on learned patterns, enabling the system to adapt to different recovery objectives without increasing operational complexity.
3Productivity
If reinforcement learning model is used to generate backup policies, then resource conflicts between backup and IO tasks are avoided, but computational cost and model training time increase
Solution Approach 1:
The reinforcement learning model is trained in advance using historical system data to learn optimal backup policies. This preliminary training phase allows the model to internalize patterns and relationships, so that during actual operation, it can quickly generate policies without requiring real-time computation, thus reducing operational time loss while maintaining high resource allocation efficiency.
Solution Approach 2:
The patent employs a two-phase approach where extensive computational resources are allocated during the offline training phase, and minimal computational resources are required during online policy generation. This partial action strategy concentrates the computational burden in advance, allowing efficient real-time operation with reduced time loss.
Data Source
AI summary
Embodiments of the present disclosure provide a method, an electronic device, and a computer program product for backing up data. The method may include acquiring a system state of a storage system according to a data backup task. The method further includes generating, by a reinforcement learning model, a plurality of candidate backup policies based on the system state. The method further includes determining a plurality of reward values for the plurality of candidate backup policies based on amounts of resources occupied by an IO task in a plurality of time periods corresponding to the data backup task. The method further includes selecting a target backup policy based on the plurality of reward values. The method further includes performing the data backup task using the target backup policy.