ZNS Memory Subsystem Shuttle Buffers for Scalable Zone Writes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing memory sub-systems face challenges in efficiently supporting a large number of zones/regions due to resource limitations, including buffer capacity, power protection, and scalability issues, particularly in managing complex address translations and media management operations.
Innovation Solution
Implementing a memory sub-system controller that uses shuttle buffers to directly store data of matching unit size in memory components, bypassing traditional buffers, and utilizing large indirection units for efficient storage and management of data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional buffers are used to store data before programming to memory components, then data can be accumulated and managed, but buffer capacity requirements increase and scalability is limited
Solution Approach 1:
The patent extracts the data accumulation function from traditional buffers by implementing a bufferless write path where the host system directly programs data to memory components. This eliminates the need for large buffer capacities while maintaining the ability to handle numerous zones/regions through direct host-initiated programming operations.
Solution Approach 2:
The host system performs preliminary data preparation and packaging before programming to memory components. By pre-organizing data in the host's buffer memory and preparing it for direct programming, the system eliminates the need for additional buffers in the memory sub-system while maintaining efficient data management for large numbers of zones.
2Ease of operation
If multiple intermediate buffers are used for data storage and translation, then data management flexibility improves, but device complexity and resource consumption increase
Solution Approach 1:
The patent removes intermediate buffers from the data path by implementing direct programming from the host to memory components. This extraction of the buffering function reduces device complexity while maintaining data management flexibility through host-side buffer memory and direct programming interfaces.
Solution Approach 2:
The host system's buffer memory acts as an intermediary for data storage before programming. Instead of requiring multiple intermediate buffers within the memory sub-system, the host's buffer memory serves as the mediator, simplifying the overall buffer hierarchy while maintaining data management flexibility.
3Productivity
If traditional buffer-based data storage is used, then data can be accumulated to optimal sizes, but write operations become sequential and throughput is limited
Solution Approach 1:
The host system performs preliminary data packaging and preparation in its buffer memory before initiating programming operations. This preliminary action allows the host to prepare data in optimal sizes for direct programming, eliminating sequential accumulation delays and enabling parallel write operations to multiple memory components simultaneously.
Solution Approach 2:
The patent enables continuous programming operations without interruption for data accumulation. By implementing direct programming from the host to memory components, the system maintains continuous useful action throughout the write process, eliminating the stop-start nature of buffer-based accumulation and significantly improving throughput.
Data Source
AI summary
The disclosure configures a memory sub-system controller to operate a memory sub-system using one or more shuttle buffers. The controller receives, from a host, a request to program a collection of data, the collection of data having a size corresponding to an individual unit size of an individual component of a set of memory components. The controller places the collection of data in an individual shuttle buffer of one or more shuttle buffers and stores the collection of data from the individual shuttle buffer to the individual component of the set of memory components without waiting for additional data to be received from the host.