UOW-Caching and new ways to improve response time (Paper)

Caching and New Ways to Improve Response Time
Guson Kuntarto (2636177)

Abstract
This essay will be focused on how a web cache proxy server work that can improve the
end-user response time. To justify there is a significant improvement using a web cache
proxy server, an example from recent study about web caching technology will be used.
This is done by comparing the mean of response times that using a web cache and
without using a web cache.

1. Introduction
In the past few years, Internet usage especially World Wide Web (www) has been
increased exponentially [3]. This forces Internet service providers (ISP) to provide
bandwidth in order to satisfy the end-users. Because often end-users request large amount
of files from servers and send them over the Internet, without enough bandwidth that
available on the network the traffic congestion may be happened. Moreover, it may lead
to the increasing number of response times from end-users perspective. In other words,
end-users tend to wait a bit longer to be able to get information that they needed.

Alternatively, implementing a web cache proxy server could be a better solution.
Configurationally, it is placed between the Internet and the end-users that act to provide
services between end-users and the Internet by reducing the number of requests sent
across the Internet to the origin servers [1]. So, it has to provide a large amount of
dedicated temporary storage to serve users. In addition, it must contain an up-to-date data
or fresh data from the popular web sites that frequently accessed by end-users [1]. By
implementing a web cache proxy server Internet bandwidth usage can be saved and the
end-user response time can be reduced [1].

In the second section of this essay will be explained caching technology terminology. In
third section, it will explain how the web cache proxy server works. Then, some

1

examples regarding to response time will be presented and compared in the section four.
Finally, conclusion will be drawn in the last section.

2. Terminology
“Cache” means to store of recent retrieved information for further reference [2]. Thus, a
web cache is a collection of locally stored of web objects such as web pages, embedded
and other types of digital objects retrieved over the Internet using the hypertext protocol
(HTTP) [1] [5]. It is maintained by a web browser and used for a private or not shared
[2]. Whereas, a web cache proxy server is essentially a shared web cache. A cache’s
program running on a dedicated server that archives and returns documents frequently
requested by a group of web clients such as squid [5].

Response time can be defined as the amount of latency that the end-user experiences
when requesting a web objects. This measurement starts when the web browser initiates a
request and completes when the server responds with the data [1].

3. How does the web cache proxy server work?
3.1. A web cache proxy server topology

Computer

Computer

PROXY ORIGIN
SERVER SERVER

Computer

Internal
Network

Figure 3.1 – A simple topology of web cache proxy server

The web cache proxy server usually placed between the internal network and Internet as
it seen on figure 3.1. It plays important rules to copy a web objects close to the users and
maintain the freshness of the web object, so that user do not have to get the objects from
the origin servers [5]. By returning local copies of objects whenever possible, a caching

2

proxy can improve response times; reduce network traffic and the effective bandwidth
available to end-users [5]. Caching proxies typically implement a cache replacement
policy including Least Recently Used (LRU), Least Frequently Used (LFU), and Hyper-
G to determine when web objects have to be removed from their archive [3][5].
Basically, LRU policy works to remove the documents that have not been accessed for
the longest time whereas LFU policy acts to delete the documents that are least frequently
asked by end-users [4]. Finally, Hyper-G performs by removing the documents, which
are LFU, and if it is found there are two items that have identical LFU, then it will delete
the one that is LRU. And if it is found two documents that have identical LFU and LRU,
Hyper-G policy will perform task to remove one that is larger [4].

3.2. Scenario of web cache proxy server

WEB CACHE
PROXY SERVER

Request Cache Miss

CLIENT

WEB SERVER

Cache Hit Store/ access History

CACHE VERSIONING

Figure 3.2 – scenario of web cache proxy server [6]

There are three scenarios that involve in web caching technology, including: cache-hit,
and cache-miss. Cache hit happen when the request from the users found the fresh copy
of data on the cache and sends it directly to the users [2]. Whereas, cache-miss happen
when the request from the users could not find the fresh copy of data that is requested on
the cache. The next task is proxy server will get the copy from the origin server, save it
on the cache for further usage and send it to the users [2]. From user perspective, as a
result of cache-hit, the response time is faster than when cache-miss is happened, because
when the cache-miss happened the proxy server has to perform task to find the fresh data
to origin server and send it to the users.

3

3.3. Web caching technique
The problem that would be faced in web cache technology is the cache interconnected to
the other cache. The web caching technique has been introduced including: Internet
Cache Protocol (ICP) and Cache Array Routing Protocol (CARP) [4]. ICP performs
using three basic messages: ICP Query, ICP_HIT and ICP_MISS for inter-cache
communication between cache and cache [4]. CARP performs HASH function to maps
string into integer [4]. By performing CARP, mirroring effect that occurs in ICP can be
solved [4].

4. Response time comparison

Figure 3.3 – Graphic cumulative distribution of request versus response time [7]

Recent study about web caching technology by Davison [7] uses NCS (Network Cache
Simulator) as a tool to get the response time accurately has been done. The experiment
was using a model simple topology of web cache proxy server, where the server attached
to the Internet with bandwidth 1,544 Mbps (T1) and the network latency is equal to 50
ms second, and 30 ms per-request overhead, whereas the client is connected to bandwidth
27,4 kbps with 200 ms of latency and requested 10000 Byte of files shows that in non-
persistence connection, the mean of response time without using a web cache proxy
server was equal to 5.5 second and with using a web cache proxy server the mean of
response time equal to 4.6 second [7]. This means there is a significant improvement that
equal to 0.9 second of response time using a web cache proxy server as it shows on the
figure 3.3. In addition, using persistent connections improves the mean of response time
in the condition where a web cache not in used drop to 5.1 second whereas with a web

4

cache the mean of response time drops to 2.4 second [7]. This means there is a significant
improvement that equal to 1.7 second of response time using a web cache proxy server in
persistent condition. Thus, using a web cache proxy will improve response time
significantly.

5. Conclusion
In conclusion, implementing a web proxy server is essential as it can reduce response
time. This because a web proxy server serves the client by placing a popular copy of web
objects as near as possible to the users. So, every time user requests data to the proxy
using protocol HTTP, proxy servers can response immediately.

Reference:
[1] IMimic, 2001, “A primer on web caching technology and benefits”, p. 1- 5, [Online]
Available: http://www.imimic.com/documents/WPBackgroundTechBenefits.pdf accessed:
04/04/2005.
[2] Dutkiewicz E., 2005, “3rd Week lecture materials of content servers and caching
technologies”, School of Electrical, Computer, and Telecommunication Engineering, University
of Wollongong, p. 4- 40.
[3] Dutkiewicz E., 2005, “4th Week lecture materials of content servers and caching
technologies”, School of Electrical, Computer, and Telecommunication Engineering, University
of Wollongong, p. 4- 40.
[4] Wooster R. P., 1996, “Optimizing response time, rather than hit rates of www proxy caches, ”
(Master of Science, Virginia Polytechnic Institute and State University), [Online] Available:
http://scholar.lib.vt.edu/theses/available/etd34131420119653540/unrestricted/etd.pdf accessed:
04/07/2005.
[5] Powell J., 1997, “Web Caching: questions and answers,” p. 1- 3, [Online] Available:
http://scholar.lib.vt.edu/digilib/reports/dlcachetalk.pdf accessed: 04/07/2005.
[6] Yang J. and Li X, 2000, “Pachyderm: The web proxy that never forgets,” p. 1- 12, [Online]
Available: www.cs.wisc.edu/~junfengy/submit/Pachyderm.pdf accessed: 04/06/2005.
[7] Davison B. D., 2001, “NCS: Network and cache simulator An Introduction’, p. 1- 13, [Online]
Available: http://www.cse.lehigh.edu/~brian/pubs/2001/dcs-tr-444/dcs-tr-444.pdf accessed:
04/07/2005

5

UOW-Caching and new ways to improve response time (Paper)

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to UOW-Caching and new ways to improve response time (Paper) (20)

Recently uploaded (20)

UOW-Caching and new ways to improve response time (Paper)