Host Write Scheduling for Die-Busy Memory Reclaim Groups
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data memory systems face issues with die busy and reduced parallelism of input/output operations due to simultaneous write requests targeting the same die, leading to performance degradation.
Innovation Solution
Implementing a host with a scheduler that divides write requests into different request sets based on reclaim group information, ensuring that requests for the same reclaim group are distributed across multiple dies, thereby preventing die busy and improving input/output parallelism.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If write requests are processed without distribution across dies, then implementation complexity is reduced, but input/output parallelism deteriorates due to die busy conflicts
Solution Approach 1:
The patent segments write requests into different request sets based on die information, distributing requests targeting the same die across multiple time slots. This segmentation resolves the contradiction by organizing requests in a structured manner that enables parallel processing without requiring complex dynamic arbitration mechanisms.
Solution Approach 2:
The patent implements periodic action by cycling through different request sets in a predetermined sequence. When a die is busy, the system periodically switches to another request set that can be processed by available dies, then returns to the original request set when the die becomes free. This periodic switching mechanism improves I/O parallelism while maintaining simple request management.
2Productivity
If write requests are distributed across multiple request sets, then input/output parallelism is improved, but device complexity increases due to scheduler requirements
Solution Approach 1:
The scheduler segments requests into numbered request sets (e.g., request set 0, request set 1) based on die information from write requests. Each request set contains requests that can be processed without die conflicts. This segmentation approach improves I/O parallelism while keeping the scheduler structure simple and manageable.
Solution Approach 2:
The scheduler performs preliminary action by pre-organizing write requests into different request sets before processing. By categorizing requests according to die information in advance, the system eliminates the need for complex real-time arbitration, achieving improved parallelism with minimal scheduler complexity.
3Speed
If simultaneous write requests are allowed on the same die, then processing speed is improved, but reliability deteriorates due to die busy conflicts and performance degradation
Solution Approach 1:
The patent applies periodic action by implementing a cyclic scheduling mechanism that switches between request sets based on die availability. When a die becomes busy with a write operation, the system periodically transitions to process other request sets using different dies, then returns to the original request set when the die is free. This maintains high processing speed while ensuring operation stability by avoiding die busy conflicts.
Solution Approach 2:
The system performs preliminary action by checking die information and organizing requests into appropriate request sets before initiating write operations. This preliminary organization ensures that simultaneous write requests are automatically directed to different dies or time slots, maintaining both high processing speed and operational reliability without requiring complex runtime arbitration.
Data Source
AI summary
A host is coupled with a memory system supporting a flexible data placement function. The memory system includes a memory device including K dies. The K dies are divided into L reclaim groups, and each reclaim group includes at least one die. K and L are positive integers greater than or equal to 2. The host includes a user layer and a scheduler. The user layer is configured to send N write requests each corresponding reclaim group information. N is a positive integer. The scheduler is configured to receive the N write requests, and place the N write requests in M request sets. Write requests carrying the same reclaim group information are placed in different ones of the request sets. M is a positive integer less than or equal to N.