Direct-Address Storage Architecture for Shorter Data Access Paths
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing storage systems have excessively long data access paths due to the need for data access requests to pass through a storage service layer and computing resources, leading to prolonged data access times.
Innovation Solution
A storage system architecture that allows direct forwarding of access addresses to hard disks without translation, utilizing computing nodes to manage data storage and storage nodes to handle data access requests, thereby shortening the data access path.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data access requests pass through the storage service layer and computing resources for address translation, then data can be stored and accessed with management capabilities, but the data access path becomes excessively long and data access time increases
Solution Approach 1:
The patent segments the storage system into distinct components: computing nodes that manage data and generate access addresses, and storage nodes that directly execute access operations using these addresses. This segmentation eliminates the need for centralized address translation through the storage service layer, allowing storage nodes to directly access data using access addresses generated by computing nodes, thereby shortening the access path while maintaining functional separation.
Solution Approach 2:
The patent introduces access addresses as an intermediary mechanism that enables direct communication between computing nodes and storage nodes without requiring traditional address translation through the storage service layer. The access address serves as a mediator that encapsulates the necessary location information, allowing storage nodes to directly locate and access data on hard disks without passing through multiple computational layers.
2Ease of operation
If address translation is performed through the storage service layer, then data can be accessed using logical addresses, but the data access path becomes longer and access speed decreases
Solution Approach 1:
The patent applies preliminary action by having computing nodes pre-generate access addresses for data before storage operations are executed. These access addresses are created in advance and stored alongside the data metadata, enabling storage nodes to directly use these pre-computed addresses for data access without performing translation operations at access time. This preliminary generation of access addresses eliminates the need for real-time address translation, thereby improving access speed while maintaining ease of operation.
3Productivity
If the storage service layer performs space management and address translation, then data can be organized and located, but the data access path becomes excessively long
Solution Approach 1:
The patent extracts the address translation function from the storage service layer and relocates it to the computing node, where access addresses are generated as part of the data management process. The storage service layer retains its space management capabilities but no longer performs address translation, which is now handled by computing nodes that generate access addresses. This extraction eliminates the need for address translation in the data access path, shortening it while preserving data management efficiency.
Solution Approach 2:
The patent enables storage nodes to serve themselves by directly using access addresses to locate and access data without requiring intervention from the storage service layer for address translation. The storage node autonomously parses the access address, determines the target hard disk and storage location, and executes the data access operation independently, thereby eliminating unnecessary intermediate steps in the access path while maintaining organized data management.
Data Source
AI summary
A storage system, a storage node, and a data storage method are provided. The storage system includes a plurality of computing nodes and a plurality of storage nodes, and each storage node includes a hard disk. A client writes to-be-stored data into a memory of a first computing node, and writes a log of the to-be-stored data into storage space corresponding to a first access address of a first storage node. This implements storage of the to-be-stored data.