Hit-Based Media Cache Band Cleaning for Storage Throughput
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data storage devices with media caches and next-level storage tiers optimized for sequential reads, such as those using shingled magnetic recording (SMR), face constrained throughput due to data fragmentation on the media cache or between the media cache and the main data store, with traditional first-in-first-out (FIFO) cleaning methods being inefficient in terms of computation and power consumption.
Innovation Solution
Implementing a hit-based band cleaning data storage device with a main data store organized into bands and a media cache with a hit-tracking module to prioritize band cleaning operations based on read request frequencies, thereby reducing fragmentation and improving read throughput by focusing on bands with higher hit counts and fragmentation levels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional first-in-first-out (FIFO) cleaning method is used on media cache, then the cleaning process is simple to implement, but read throughput is constrained due to data fragmentation and computational inefficiency
Solution Approach 1:
The patent changes the cleaning selection parameter from FIFO order to hit count frequency. The band cleaning module selects bands for cleaning based on hit count thresholds and fragmentation levels rather than insertion order, transforming the cleaning strategy from time-based to performance-based selection
Solution Approach 2:
The hit-tracking module performs preliminary counting of read requests for each band before the cleaning operation. This advance tracking of access patterns allows the system to identify high-priority bands for cleaning beforehand, enabling optimized cleaning sequences that prevent future fragmentation
2Reliability
If band cleaning is performed on all bands regardless of access pattern, then data fragmentation is reduced uniformly, but power consumption and computational resources are wasted on low-access bands
Solution Approach 1:
The patent applies different cleaning priorities to different bands based on their local access characteristics. Bands with high hit counts and fragmentation levels are selected for cleaning while low-access bands are left untouched, creating a non-uniform cleaning strategy tailored to each band's actual usage pattern
Solution Approach 2:
Instead of cleaning all bands uniformly, the system performs partial cleaning only on bands that exceed hit count thresholds and show fragmentation. This selective approach applies cleaning action only where necessary, avoiding waste of computational resources and power on bands that don't require cleaning
3Quantity of substance
If media cache is kept fully utilized to maximize storage capacity, then storage efficiency is improved, but read performance degrades due to increased data fragmentation
Solution Approach 1:
The system continuously monitors hit counts and fragmentation levels as feedback metrics. Based on this feedback, the band cleaning module dynamically adjusts which bands to clean and when, creating a closed-loop control system that maintains optimal balance between cache utilization and read performance
Solution Approach 2:
The cleaning strategy is made dynamic rather than static. The system adapts cleaning priorities based on changing access patterns and fragmentation levels, allowing the media cache to maintain high utilization while dynamically preventing fragmentation in high-access bands
Data Source
AI summary
A data storage device includes a media cache and a main data store optimized for sequential reads and organized into bands. When the data storage device receives a read request from a host computing system, the requested data may be fragmented across the media cache and the main data store, causing constrained read throughput. Band rewrite operations to improve read throughput are selected based on a hit tracking list including a hit counter associated with each band on the main data store. The hit counter tracks the number of times a host computing system has requested data in logical block addresses corresponding to the various bands. The data storage device may select bands for band rewrite operations based on the number of hits in the associated hit tracking counters.