Cache Page Flushing Algorithm Using Access Frequency and Time
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current page replacement policies in storage systems, such as the LRU algorithm, fail to consider access frequency, leading to inefficient flushing of pages, where high-frequency pages are evicted by low-frequency pages, resulting in wasted cache resources and unfairness among volumes, and lack of relevance between shadow and cached pages, causing undesirable promotions and loss of historical information.
Innovation Solution
Implementing a page flushing algorithm based on access time and frequency, organizing pages into queues within volumes, and using a weighted unaccessed time length to determine which pages to flush, ensuring fairness and maintaining cache efficiency by prioritizing high-frequency pages and avoiding stale pages.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If LRU algorithm is used for page replacement, then cache management is simple, but high-frequency pages are evicted by low-frequency pages leading to wasted cache resources
Solution Approach 1:
The patent introduces access frequency as an additional parameter to the traditional LRU time-based replacement policy. By combining time (LRU) and frequency parameters, the system determines page replacement priority more accurately, retaining high-frequency pages even if they are less recently accessed, thus improving cache efficiency without significantly increasing complexity
Solution Approach 2:
The patent creates a composite replacement policy that combines LRU (time-based) and access frequency-based mechanisms. This composite approach integrates multiple dimensions (time and frequency) into a unified replacement strategy, allowing the system to leverage both recent access patterns and historical access frequency to make better replacement decisions
2Ease of operation
If pages are flushed based solely on least recently used criterion, then eviction policy is straightforward, but fairness among volumes is compromised
Solution Approach 1:
The patent applies different weighting factors to different volumes based on their access frequency characteristics. Each volume can have its pages evaluated with locally adapted parameters, allowing the system to maintain fairness across diverse workloads while keeping the overall eviction policy manageable through localized adjustments rather than global complexity
3Loss of energy
If shadow pages and cached pages are managed independently, then management overhead is low, but relevant promotions are lost and historical information is discarded
Solution Approach 1:
The patent merges the management of shadow pages and cached pages into a unified structure where both types of pages share the same access frequency tracking and replacement priority calculation. This integration allows the system to leverage historical information from shadow pages to inform caching decisions, preventing loss of valuable historical data while maintaining efficient management through a single coordinated mechanism
Data Source
AI summary
Techniques provide a storage management mechanism for flushing a page out of cache. The techniques involve obtaining recently used (RU) information of cache. The cache includes volumes, each of which includes a plurality of pages. The RU information indicates at least part of the volumes, and an unaccessed time length of pages in the at least part of the volumes exceeds a threshold time length. The techniques further involve: determining, based on a total number of pages to be flushed out of the cache and the RU information, a number of pages to be flushed out of the cache from the at least part of the volumes. The techniques further involve flushing, based on the determined number, pages in the at least part of the volumes. Accordingly, cache efficiency can be increased, and the overall performance of the disk array can be increased accordingly.