Storage System Local Data Recovery via Controller Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In network-connected storage systems, data recovery processes are delayed and host performance is unstable due to network bottlenecks, especially when using Drive Box (FBOF) configurations, where data redundancy methods exacerbate performance issues and reliability concerns.
Innovation Solution
A storage system architecture that includes multiple storage units and a calculator connected via a network, where data redundancy is managed by the storage controller and processor to reduce network load by selectively executing data recovery processes within the storage unit, using Chunk Groups and parity methods to minimize network transfer during failures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data recovery is performed by the storage controller via network, then data redundancy can be achieved, but network transfer amount increases and causes performance bottleneck
Solution Approach 1:
The patent segments the data recovery function into two parts: local recovery within the storage unit using its processor, and remote recovery via storage controller. This segmentation allows most recovery operations to be handled locally without network transfer, reducing network load while maintaining redundancy capability.
Solution Approach 2:
The storage unit's processor acts as an intermediary that can perform data recovery locally before transmitting results to the storage controller. This intermediary capability reduces the need for network-mediated recovery operations, decreasing network transfer amount while preserving data redundancy.
2Reliability
If data recovery is performed by reading large amount of data via network, then data can be recovered from failed drive, but data recovery process is delayed
Solution Approach 1:
The storage unit performs preliminary data recovery actions locally using its processor before involving the storage controller or network. By preparing and recovering data locally in advance, the system reduces the time required for full data recovery and minimizes network involvement.
Solution Approach 2:
The system skips the network transfer step for data recovery by enabling local processing within the storage unit. This allows the recovery process to rush through the critical path without network bottlenecks, significantly reducing data recovery time.
3Quantity of substance
If FBOF performs data redundancy, then storage capacity is increased, but FBOF becomes performance bottleneck
Solution Approach 1:
The storage unit is equipped with its own processor that can independently perform data recovery and redundancy operations without requiring constant intervention from the storage controller or network. This self-service capability maintains storage capacity while eliminating the performance bottleneck by handling operations locally.
4Reliability
If data recovery is performed via network, then data can be restored, but host performance becomes unstable
Solution Approach 1:
The patent extracts the data recovery operation from the network path and places it within the storage unit's local processing capability. This extraction removes the source of performance instability from the host system, allowing data restoration to proceed without affecting host performance.
Data Source
AI summary
A storage system includes one or more storage units, and a calculator. Each of the one or more storage units has plural physical storage devices, and a processor. The calculator has a controller that processes data inputted to and outputted from the storage unit by the processor. The storage system subjects the data to redundancy and stores the redundant data, and when a failure in which the data related to a read request cannot be read from one of the physical storage devices has occurred, reads the data from the readable physical storage device, recovers the data related to the read request, and transmits the recovered data to a request source of the read request. A process for recovering the data related to the read request from the read data can be selectively executed by the controller of the calculator and the processor of the storage unit.