Storage Cache Tiering Based on Disk Processing Capability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current storage systems face performance issues due to processing capability differences among hard disks, leading to cache page resource preemption and reduced quality of service, where slow disks hinder fast disks and cause a 'bottleneck effect', as existing cache partition technologies do not effectively manage these differences.
Innovation Solution
Divide the cache into multiple tiers based on the processing capabilities of hard disks, allocating page quotas accordingly, and write data into corresponding hard disks when a preset proportion of dirty data is reached, ensuring independent operation of each tier and preventing slow disks from affecting fast disks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If cache partition technology is used to divide cache resources into multiple cache partitions according to services on a host, then cache resources can be allocated to different services, but a page resource preemption phenomenon still occurs in each cache partition, and the performance of a service corresponding to a cache partition is affected by the slowest back-end disk
Solution Approach 1:
The patent divides the cache into multiple tiers (first cache tier, second cache tier, etc.) based on the processing capabilities of back-end disks. Each cache tier is independently managed and corresponds to disks with similar performance characteristics. This segmentation prevents fast disks from being hindered by slow disks, as each tier operates independently with its own page resources that cannot be preempted by other tiers.
Solution Approach 2:
The patent applies local quality by assigning different cache tiers to different service types based on their performance requirements. High-performance services are assigned to cache tiers associated with fast disks, while lower-performance services are assigned to cache tiers with slower disks. This ensures that each service receives cache resources matched to its actual needs, preventing high-performance services from being degraded by slow disk operations.
2Adaptability or versatility
If various types of hard disks with different rotational speeds and throughput are mixed in the storage system, then storage system flexibility is improved, but processing capability differences cause slow disks to occupy cache page resources and affect the processing speed of fast disks
Solution Approach 1:
The patent segments the cache into multiple tiers based on back-end disk processing capabilities. Each cache tier is associated with disks of similar performance levels, creating isolated processing domains. This segmentation ensures that slow disks cannot occupy or preempt cache page resources allocated to fast disks, as each tier manages its own page resources independently.
Solution Approach 2:
The patent applies local quality by matching cache tiers to specific disk performance characteristics. Each cache tier is optimized for the processing capabilities of its associated disks, ensuring that fast disks receive appropriate cache resources without being hindered by slow disks. This local optimization maintains high processing speeds for each disk type while supporting mixed disk configurations.
3Productivity
If the cache size is enlarged to improve read operation performance, then the cache hit ratio can be raised, but when the cache is enlarged to a certain extent, a performance improvement of the read operation and an increase in the size of the cache are disproportionate
Solution Approach 1:
The patent divides the cache into multiple tiers based on back-end disk processing capabilities rather than simply increasing overall cache size. This segmentation allows the system to optimize cache utilization by matching cache resources to actual disk processing speeds, preventing waste of cache resources on slow disks while maintaining high cache hit ratios for fast disks without requiring excessive total cache capacity.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
The present invention provides a method and a system for controlling quality of service of a storage system, and a storage system. The method includes: collecting statistics about processing capability information about hard disks in the storage system and obtaining processing capabilities of the hard disks according to the processing capability information; dividing a cache into multiple tiers according to the processing capabilities of the hard disks; and writing, for a cache of a tier in which dirty data reaches a preset proportion, data in the cache of the tier into a hard disk corresponding to the cache of the tier. The present invention avoids a phenomenon of preempting page resources in the cache.