Storage System Task Allocation and Dynamic Polling for Accelerator Resources
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing storage systems face challenges in efficiently allocating tasks of different types to multiple accelerator resources, such as QAT cards, and determining appropriate polling intervals for task execution.
Innovation Solution
A method that allocates tasks based on their types to various accelerator resources and determines a polling interval for each resource based on the average task size of the allocated tasks, ensuring balanced task distribution and optimized execution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If tasks of different types are allocated to multiple accelerator resources without considering task size, then accelerator resources can be utilized, but polling intervals cannot be optimized leading to unnecessary polls and increased latency
Solution Approach 1:
The patent segments tasks into different groups based on their task sizes (small, medium, large tasks) and allocates them to different accelerator resources. This segmentation enables the system to determine different polling intervals for different task groups, optimizing the balance between resource utilization and polling efficiency.
Solution Approach 2:
The patent implements dynamic polling intervals that adapt to the characteristics of allocated tasks. By determining polling intervals based on average task size of allocated tasks, the system dynamically adjusts polling frequency to match actual task processing needs, reducing unnecessary polls while maintaining efficient task completion notification.
2Device complexity
If accelerator resources are allocated without considering task types, then resource allocation is simple, but task distribution becomes unbalanced leading to suboptimal system performance
Solution Approach 1:
The patent applies local quality by allocating different types of tasks to different accelerator resources based on their characteristics. Small tasks are allocated to one set of accelerators, medium tasks to another, and large tasks to a third set. This localized optimization ensures that each accelerator resource operates in its optimal performance range, improving overall system productivity.
Solution Approach 2:
The patent changes the allocation parameter from simple round-robin or random distribution to task-size-based classification. By using task size as the allocation parameter, the system achieves balanced task distribution across accelerator resources, optimizing processing efficiency without requiring complex allocation algorithms.
3Ease of operation
If a fixed polling interval is used for all accelerator resources, then scheduling is simple, but tasks with different sizes experience suboptimal polling efficiency
Solution Approach 1:
The patent transitions from fixed polling intervals to dynamic polling intervals that are determined based on the average task size of tasks allocated to each accelerator resource. This dynamic adjustment optimizes polling efficiency for different task types while maintaining relatively simple scheduling logic through automated interval calculation.
Solution Approach 2:
The patent implements a feedback mechanism where the polling interval for each accelerator resource is determined based on the average task size of its allocated tasks. This feedback loop allows the system to automatically adjust polling intervals to match actual task processing characteristics, optimizing polling efficiency without manual intervention.
Data Source
AI summary
Embodiments of the present disclosure relate to a method, an electronic device, and a computer program product for managing a storage system. The method includes: based on respective task types of a plurality of tasks to be executed, allocating the plurality of tasks to a plurality of accelerator resources in a storage system for processing; at least for a first accelerator resource in the plurality of accelerator resources, determining a first polling interval based on an average task size of a first group of tasks allocated to the first accelerator resource; and scheduling the execution of the first group of tasks at the first accelerator resource at the first polling interval. The embodiments of the present disclosure can optimize the scheduling of the tasks to be executed on the plurality of accelerator resources, thereby optimizing system performance.