Storage Node Retry Interval Control for Latency Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In storage systems with multiple clients and nonvolatile storage devices, commands may not be executed in a timely manner due to busy states, leading to repeated transmissions and increased latency and data traffic congestion.
Innovation Solution
A method is introduced where a connection unit determines a retry interval based on the operation status response from a target node module, allowing for optimized re-transmission of access requests, thereby reducing unnecessary retries and prioritizing packets to non-busy node modules.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If commands are repeatedly transmitted to a busy storage device, then the commands may eventually be executed, but data traffic congestion increases and latency increases
Solution Approach 1:
The connection unit checks the operation status of the storage device before transmitting commands. By performing this preliminary check, the system avoids transmitting commands to busy devices, preventing unnecessary retries and reducing latency while ensuring commands are only sent when execution is likely to succeed
2Reliability
If commands are repeatedly transmitted to a busy storage device, then the commands may eventually be executed, but data traffic congestion increases
Solution Approach 1:
The connection unit performs a preliminary status check before command transmission to determine if the storage device is busy. This prevents unnecessary command retries during busy periods, reducing the quantity of redundant data traffic in the network while maintaining reliable command execution when devices are available
3Device complexity
If the timing of busy state end cannot be predicted, then simple retry mechanisms can be used, but repeated transmissions occur increasing latency
Solution Approach 1:
The storage device provides operation status information (busy or idle) as feedback to the connection unit. This feedback mechanism allows the connection unit to make informed decisions about command transmission timing, avoiding blind retries during busy states and significantly reducing latency compared to simple fixed-interval retry mechanisms
Data Source
AI summary
A storage system includes a plurality of nodes, each of the nodes including one or more node modules each of which includes a nonvolatile storage, and a connection unit directly connectable to at least one of the nodes. The connection unit is configured to transmit an access request or an inquiry directed to a target node module, determine a length of an interval before re-transmitting the access request or the inquiry, based on a response indicating an operation status of the target node module, which is returned by the target node module in response to the access request or the inquiry, and re-transmits the access request or the inquiry after the interval of the determined length has passed.