You are on page 1of 2

Design and Performance Analysis of the Web Caching

Algorithm for Performance Improvement of the


Response Speed

Yun Ji Na1 and Il Seok Ko2,*

1
Division of Internet Software Engineering, Honam University, Gwangju, Korea
yjna@honam.ac.kr
2
Corresponding Author, Division of Computer and Multimedia, Dongguk University, 707
Seokjang-dong, Gyeongju-si, Gyeongsangbuk-do, 780-714 Korea
isko@dongguk.edu

Abstract. Until now, various web caching replacement algorithms related to


this were researched. An aggravation of web object heterogeneity more
frequently generates a replacement of a web object, but cannot reflect enough
all kinds of characteristics of web object in these traditional replacement
techniques. Especially size heterogeneity of an object has an influence on
performance of a web caching seriously. Therefore, a study on web caching
algorithm with size heterogeneity of an object is required.
In this study, we proposed a web object replacement algorithm with
heterogeneity of a web object for performance improvement of the response
speed. The algorithm is designed with a divided scope that considered size
reference characteristic and reduced size heterogeneity on web object. The
performance of the algorithm is analyzed with an experiment. With the
experiments results, the algorithm is compared with previous replacement
algorithms, and its performance is confirmed with 10%-20% elevation of
object-hit ratio and an improvement of response speed.

Keywords: size variation, object heterogeneity, caching, web object.

1 Introduction
Generally we use web cache in order to improve the performance of the web base
system. Cache performance depends heavily on a replacement algorithm, which
dynamically select a suitable subset of objects for caching in a finite cache space.
Until now LRU, LRUMIN, SIZE and various web caching replacement algorithms
related to this were researched. An aggravation of web object heterogeneity more
frequently generates a replacement of a web object, but cannot reflect enough all
kinds of characteristics of web object in these traditional replacement techniques.
Especially size heterogeneity of an object has an influence on performance of a web
caching seriously. Therefore, a study on web caching algorithm with size
heterogeneity of an object is required.
*
Corresponding Author.

T.-J. Cham et al. (Eds.): MMM 2007, LNCS 4352, Part II, pp. 686 693, 2007.
Springer-Verlag Berlin Heidelberg 2007
Design and Performance Analysis of the Web Caching Algorithm 687

In this study, we proposed a web object replacement algorithm with heterogeneity


of a web object. The algorithm is designed with a divided scope that considered size
reference characteristic and reduced size heterogeneity on web object. The
performance of the algorithm is analyzed with an experiment. The reference
characteristics of size to occur by user reference characteristics are investigated
through log analysis of a web base system in the experiment. The cache scope is
divided with the results of the analysis, and the replacement algorithm is tested on
those divided scope. With the experiment results, the algorithm is compared with
previous replacement algorithms, and its performance is confirmed with 10%-20%
elevation of object-hit ratio and an improvement of response speed.

2 Related Work
Table 1 presents a comparison of various caching methods. As noted in Table 1,
studies on the existing caching method have focused on the improvement of the
performance of caching through an effective replacement of objects.

Table 1. Caching method

Items Classification Features


Applicable or available for the field of Internet
LRU caching with a peculiarity of object size
Traditional variability
Method To prepare an affordable space for new
LRU Extensions incoming Internet objects, this procedure
deletes from the latest unused objects
SIZE Replace the largest objects
This method enables preventing multiple small
Based on a documents from being replaced by applying
key LRUMIN the LRU only to larger objects rather than a
new one in the application of LRU to Internet
caching
This method enables deleting an object of the
GD(Greedy Dual)-
smallest size and the lowest transmission cost
Size
from the cache memory
Based on a
This method enables selecting a deleted object
cost
using a correlative function weighted with
Bolot & Hoschka
transmission cost, object size, and the last time
of request

Web-caching algorithms to relieve network congestion and access latency have


proliferated in recent years. The following descriptions indicate how each algorithm
selects a victim to purge from the cache.
1) LRU (least recently used): Removes the least recently referenced object first. LRU
is a algorithm to replace an unused object in storage scope so that a new object gets a
storage space.
2) LFU (least frequently used): Removes the least frequently referenced object
first[4]. LRU and LFU have applied the traditional replacement algorithm to web

You might also like