Multicast SSD Controller Reduces Command Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current communication methods between hosts and SSDs using peer-to-peer protocols lead to increased latency due to the need for command propagation and acknowledgment between storage devices, requiring additional memory space and processing time.
Innovation Solution
Implementing a multicast communication protocol that allows a host to transmit commands to multiple SSDs simultaneously, with each SSD determining its role as either a master or replica based on logical addresses, enabling parallel processing and reduced latency through shared submission and completion queues.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If peer-to-peer communication method is used between host and SSDs, then commands can be transmitted to multiple SSDs, but latency increases due to command propagation and acknowledgment requirements
Solution Approach 1:
The system segments the command processing role into two distinct parts: the master SSD that receives and processes commands from the host, and replica SSDs that only execute commands and send acknowledgments. This segmentation allows the master SSD to handle command interpretation and decision-making, while replica SSDs focus solely on execution, reducing overall processing latency and improving throughput.
Solution Approach 2:
The master SSD acts as an intermediary between the host and replica SSDs. It receives commands from the host, determines which replica should execute each command based on logical addresses, and coordinates the acknowledgment process. This intermediary role eliminates the need for direct host-to-replica communication and reduces command propagation latency.
2Adaptability or versatility
If peer-to-peer communication method is used between host and SSDs, then multiple SSDs can be accessed, but additional memory space is required for command propagation and acknowledgment
Solution Approach 1:
The system merges the submission queue and completion queue resources at the master SSD level, allowing all replica SSDs to share these memory structures through the master's coordination. Instead of each SSD requiring separate queues, the master SSD manages a single set of queues that serve the entire system, significantly reducing total memory space requirements while maintaining multi-SSD access capability.
3Productivity
If each SSD processes commands independently, then processing parallelism is achieved, but load balance among SSDs deteriorates
Solution Approach 1:
The system assigns different functional qualities to different SSDs based on their roles. Master SSDs are equipped with command processing and decision-making capabilities, while replica SSDs are optimized for command execution and acknowledgment. This local quality differentiation allows parallel processing of commands while maintaining centralized load distribution, ensuring balanced workload across the storage system.
Data Source
AI summary
A storage device includes a memory device; and a controller configured to fetch a command from a host, the command indicating a logical address, process the command based on the logical address, and receive, from a first replica storage device, an acknowledgment signal indicating that the command has been processed by the first replica storage device.