RAID Write-Back Scheduler for Load Balancing and Buffer Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
RAID apparatuses face inefficiencies in write-back processes due to heavy loads on specific hard disk drives and buffer shortages, leading to suspended write-backs and interference with other processes, especially when supporting RAID4, RAID5, or RAID6 levels.
Innovation Solution
A RAID apparatus with a write-back scheduler that determines data to be written back based on usage state and buffer availability, prioritizing write-backs to balance loads and manage buffer requests, ensuring efficient execution and reducing suspensions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If write-back is concentrated on specific hard disk drives, then write-back efficiency improves, but load on those drives becomes heavy and overall write-back efficiency reduces
Solution Approach 1:
The patent implements dynamic write-back scheduling that adapts to changing system conditions. The controller monitors buffer availability, drive load states, and data access patterns in real-time, dynamically adjusting which drives receive write-back operations. This prevents static concentration of write-back loads on specific drives while maintaining high efficiency through adaptive optimization.
Solution Approach 2:
The system changes operational parameters based on system state. When buffers are available and drive load is low, write-back operations are prioritized. When buffers are scarce or drives are heavily loaded, the scheduler adjusts parameters to suspend or delay write-back operations, preventing overload and maintaining system reliability.
2Reliability
If more hard disk drives are used for RAID4, RAID5 or RAID6 levels, then redundancy and capacity improve, but more buffers are needed causing write-back suspension
Solution Approach 1:
The patent implements partial write-back action by suspending write-back operations when buffer resources are insufficient. Rather than attempting to maintain full write-back throughput regardless of buffer availability, the system performs partial write-back only when buffers are available, preventing write-back suspension while supporting RAID4/5/6 redundancy requirements.
Solution Approach 2:
The system performs preliminary buffer allocation and availability checking before initiating write-back operations. By checking buffer status in advance and preparing appropriate buffer resources, the system prevents write-back suspension caused by buffer shortages, enabling reliable operation with multiple RAID drives.
3Reliability
If other accesses to cache memory are exclusively controlled during write-back process, then write-back integrity improves, but other host processes are interfered with
Solution Approach 1:
The patent implements periodic or interrupt-based cache access control during write-back operations. Instead of maintaining exclusive control throughout the entire write-back process, the system periodically allows other host processes to access the cache memory when write-back operations are not actively executing, reducing interference time while maintaining write-back integrity through controlled access windows.
Solution Approach 2:
The system dynamically adjusts cache memory access control based on write-back operation state. When write-back is actively executing, exclusive control is maintained for integrity. When write-back is suspended or idle, the control policy dynamically changes to allow other processes, minimizing interference while preserving write-back reliability.
4Productivity
If LRU method is used for write-back scheduling, then overall performance is optimized, but free cache memory space is not efficiently increased
Solution Approach 1:
The patent changes the write-back scheduling parameter from pure LRU (Least Recently Used) to a composite parameter that includes buffer availability and cache space utilization. This allows the system to prioritize write-back operations that free cache memory space when buffers are available, rather than strictly following LRU order, thereby increasing free cache memory while maintaining performance optimization.
Data Source
AI summary
A RAID apparatus includes a plurality of recording devices, a first adaptor connected to a first interface which is connected to a high-level apparatus, a controller for controlling processing of data transmitted by the high-level apparatus, and a second adaptor that connects to a second interface connected to a plurality of recording devices. The controller has a first memory area in which said data are stored, a second memory area used for a write-back of the data in said plurality of recording devices, a write-back information control unit controlling a write-back type of data stored in the first memory area and a usage state of the second memory area, a write-back data determination unit for determination of data to be written-back based on the write-back information, and a write-back execution unit for executing a write-back of the data to be written back on the plurality of recording devices.