Browse Prior Art Database

LRU Improvement with Social Awareness Approach

IP.com Disclosure Number: IPCOM000243041D
Publication Date: 2015-Sep-10
Document File: 6 page(s) / 82K

Publishing Venue

The IP.com Prior Art Database

Abstract

In a multiple users and constanlty changing workload environment, LRU only couldn't maximizing cache's performance. It's due to limited cache resource competited by multiple peers. By introducing a generalized sticky cache, which hosts common data or data likely be accessed in the near future, we can improve cache and the application's performance. How to identify these data? The social awareness approach ia used. By analyzing user's characteristics and behaviors, data is put into sticky cache to maximize cache performance. Several data mining techniques, such as clustering, classification, association, etc, are used.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 58% of the total text.

Page 01 of 6

LRU Improvement with Social Awareness Approach

Least Recently Used (LRU) is an ancient cache replacement algorithm, which was quite useful in single user environment. However, nowadays, social-rich multiple-user applications are prevalent. In order to reach better cache performance,

we need to inject more social factors in cache replacement strategy.

A general sticky cache is introduced to tackle many social factors. The mechanism of the sticky cache is shown as follows:

In the application, there are normal cache and sticky cache work together to serve cached data.

The normal cache uses LRU as the replacement strategy to cover the temporal factor of data usage.

The sticky cache uses an index Hit Ratio Max, HRM(User coverage, Time window, User weight, ...) to decide which arrangement is best to achieve the performance goal of the application.

The system will periodically inject data to the sticky cache, based on real time HRM calculation.

1


Page 02 of 6

Design an effective and flexible sticky cache,

User grouping: divide the whole user group into several sub-groups either by business rules or data-driven methods.

Or directly use the whole group.

,In each sub-group or whole group, define a hard coverage minimal threshold, or compute a coverage score for each resource, or combine the two criterion together,

2


Page 03 of 6


Select the resources to sticky cache that
,Maximize the coverage rate summed by each coverage score by a weight, the weight can be set by business...