Networked SSD Data Routing by NAND Read Pattern
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The high cost of manufacturing SSDs with asynchronous independent plane read (AIPR) enabled for all NAND memory due to sunk costs when the feature is not universally utilized, necessitating a need for cost-effective data routing based on data retrieval types.
Innovation Solution
A master-SSD routes data to multiple client SSDs via a controller memory buffer, analyzing application data retrieval types and selectively utilizing SSDs with AIPR or non-AIPR NAND circuitry based on sequential or random data patterns, optimizing hardware costs by using AIPR circuitry only when performance benefits are maximized.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If all NAND memory is manufactured with AIPR enabled, then random read performance is improved, but manufacturing cost increases due to sunk costs for features not universally utilized
Solution Approach 1:
The patent segments the storage system into a master SSD that coordinates data placement and multiple client SSDs with different NAND configurations (AIPR-enabled and non-AIPR). This segmentation allows the system to optimize for random read performance only where needed (in AIPR SSDs) while avoiding the sunk cost of manufacturing all SSDs with AIPR capability, thereby resolving the contradiction between performance improvement and manufacturing cost increase.
2Ease of manufacture
If AIPR circuitry is selectively used based on data retrieval type, then hardware cost is reduced, but system complexity increases due to data routing requirements
Solution Approach 1:
The master SSD acts as an intermediary that manages the complexity of data routing between applications and client SSDs with different NAND configurations. It analyzes data retrieval patterns and routes random read workloads to AIPR-enabled SSDs while directing sequential reads to non-AIPR SSDs. This intermediary approach hides the routing complexity from individual SSDs and applications, allowing cost reduction through selective AIPR usage while maintaining manageable system complexity.
3Reliability
If data is routed based on retrieval type analysis, then quality of service is maintained, but processing overhead increases due to analysis requirements
Solution Approach 1:
The master SSD performs preliminary analysis of data retrieval patterns when data is initially written or accessed, classifying data as suitable for random read or sequential read operations. This preliminary classification allows the system to maintain quality of service by pre-positioning data in appropriate SSDs before actual retrieval operations, thereby reducing processing overhead during subsequent data access operations while ensuring the right data resides in the right location.
Data Source
AI summary
Instead of all the NAND memory having asynchronous independent plane read (AIPR) enabled, a system utilizes a master-SSD to route data to multiple client SSDs via a controller memory buffer (CMB) according to the underlying NAND hardware (HW) of each of the client SSDs. The master-SSD will also analyze the application data retrieval of the logical data that is to be routed. The system comprises SSDs with AIPR NAND circuitry and SSDs with non-AIPR NAND circuitry. The system further comprises the master-SSD that will route data to any of the SSDs in the system based on the data retrieval type (sequential or random). Each of the SSDs can be of a different type beyond the NAND chip, such as HW IPs or specific firmware (FW) that is tuned towards specific data routing and retrieval requirements.