Cache Admission Control Using Utility-Based Data Prioritization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Cache systems face inefficiencies in admitting data items due to limited capacity, leading to missed opportunities for time-saving and increased costs in flash-based systems, as they often admit data items on subsequent accesses rather than initial ones, missing the chance to cache items at first access.
Innovation Solution
A method for cache admission control that categorizes access requests, determines utility values based on historical or real-time data, and admits new data items into the cache in descending order of utility until capacity is reached or a negative utility is encountered, optimizing cache storage by prioritizing frequently accessed items.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If data items are admitted into cache on first access, then data access speed is improved for subsequent accesses, but cache space is consumed and flash write costs increase
Solution Approach 1:
The patent changes the admission policy parameter from uniform (admit all or admit none) to differentiated (admit based on category and utility value). By categorizing access requests and computing utility values, the system selectively admits only high-value data items, transforming the admission decision from a binary choice to a nuanced, multi-parameter decision process that balances speed improvement against write costs
Solution Approach 2:
The patent performs preliminary categorization and utility value computation before admission decisions are made. By pre-classifying access requests into categories and calculating expected utility values in advance, the system prepares admission criteria beforehand, enabling efficient real-time decisions that admit only data items with high predicted benefit while avoiding unnecessary flash writes
2Loss of energy
If data items are not admitted on first access (Lazy Adaptive Replace Cache), then flash write costs are reduced, but time-saving opportunities are missed
Solution Approach 1:
The patent transforms the单一的 admission policy into a multi-parameter decision framework by introducing category classification and utility value computation. This allows the system to selectively admit data items based on their predicted benefit, admitting high-utility items immediately (capturing time savings) while excluding low-utility items (avoiding unnecessary flash writes), thus resolving the contradiction between LARC's cost savings and missed opportunities
Solution Approach 2:
The patent performs preliminary categorization and utility assessment before admission decisions, enabling the system to identify which data items are worth caching in advance. This preliminary analysis allows the system to proactively admit high-value data items on first access, capturing time-saving opportunities that LARC misses, while still maintaining cost efficiency by excluding low-value items
3Speed
If cache admits all accessed data items, then data access speed is improved, but cache capacity is exceeded and eviction overhead increases
Solution Approach 1:
The patent changes the admission criterion from universal (admit all accessed items) to selective (admit only items in identified categories with positive utility). By introducing category filtering and utility value thresholds, the system reduces the volume of data requiring admission decisions, thereby decreasing eviction overhead while maintaining speed benefits for high-value data
Solution Approach 2:
The patent performs preliminary categorization and utility computation to pre-identify which data items warrant cache admission. This preliminary filtering reduces the burden on the eviction mechanism by preventing low-value items from entering the cache in the first place, thereby reducing eviction overhead while preserving fast access for high-utility data items
Data Source
AI summary
The technology is directed to cache admission control. One or more processors may categorize access requests for data items in a cache storage into a plurality of categories and collect information on the access requests over time. Based on the collected information, a utility value for caching data items in each category of the plurality of categories may be determined. Newly requested data items may be admitted into the cache storage in an order according to the corresponding utility values of their respective categories.