Optimizing Predictive Pre Fetching in Multi-Client

8/3/2019 Optimizing Predictive Pre Fetching in Multi-Client

http://slidepdf.com/reader/full/optimizing-predictive-pre-fetching-in-multi-client 1/27

Optimizing Predictive Prefetching in Multi-Client Single-Server Environment

Presented byPresented byNaveedNaveed AhmadAhmad

AzamAzam Khan,Khan, FaizaFaiza BibiBibi

99thth International Conference on Frontier of Inf ormationInternational Conference on Frontier of Inf ormationTechnologyTechnology

Dated:20/12/2011Dated:20/12/2011



Agenda

1. Introduction

2. Literature Review

3. Proposed Solution4. Experiment

Phase 1: Data Collection (IIU, NUST, HU)

Phase 2: Data Cleansing

Phase 3: Pr ocessing Data(Method)

Phase 4: Patter n Discovery & Patter n AnalysisEquation & Compar ison of Results



Motivation

Huge amount of research is done in semantic

web

Organization often need to access their data fast



Optimizing Predictive Prefetching in Multi

Client Single-Server Environment

Introduction

Web caching

A technique used for the client, server and proxies

to help users for fast access to the web

Web Prefetching

Process of accessing the web objects before the user

request



Optimizing Predictive Prefetching in Multi-Client

Single-Server EnvironmentPrefetching Types

Server initiated pref etching(method at server side)

Client initiated pref etching (method/ statistics at

client side)

Proxy initiated pref etching (Proxy side method)



Main Objective

The main objective of this work is the comparisons

of propose techniques with existing techniques for

enhancing previous techniques

This model navigated the more efficient results as

compared to the techniques proposed in the [12]and [13]



Literature Review

In web log mechanism server keep the session of itseach client[13]

From the Log, the probability of each web objectfor pref etching is calculated

Objective

E

nhances CachingMining Web Logs for Prediction Models in WWW

Caching and Pref etching[13]



Literature Review

When a client requests for a web page, before

accessing the web page a prediction is made foraccessing that web page[12]

By implementation and graphical analysis, results

showed that proposed model out performed theexisting[13] algorithm in Web Page Pref etching

Mechanism

Major advantage of the Sequential RankSelection algorithm is that It selects only one web

page object of a website for pref etching purposes

of user; hence consumed much less memory

space of users [12]



Literature Review

Web usage information about the pages of users

Proposed Solution

The proposed architecture consists on the followingparts

1. Client side

2. Server side

3. Page History Storage Module4. Intermediate Prediction Engine Module



Fig: 1 Framework for Multi client- Server in predictive prefetching

Proposed Solution



Experiment

Phase 1: Data Collection (IIU, NUST, HU)

Phase 2: Data Cleansing

Phase 3: Pr ocessing Data(Method)

Phase 4: Patter n Discovery & Patter n Analysis

thr ough Equations



Flow of data during Prefeching

Fig: 2 Flow of data in the proposed framework for web prefetching



P hase 1: Data Collection

The data contain the web logs information of diff erentusers requested for the resources

These are local data set of International Islamic

University, National University of S

cience andTechnology and Hazara University web servers

Data set of International Islamic University = 3000 webobject

Data set Hazara University= 3500 web object

Data set National University of Science and Technology

=4000 web object



P hase 2:Web log Data Cleansing

Cleaned the unnecessary information from the web

logs

Transf ormation of Web Logs data into Readable

The data sets taken from servers are in notepad

format The data in the notepad is seen to be raw and

unreadable. In order to change the data sets into

understandable format

We have converted the data sets of server into MS

Excel tables



P hase 3: Data Processing



P hase 4:

Pattern Discovery & Pattern Analysis

This procedure is repeated for each clients request

The percentage of each object is calculated by

the formula Percentage (Web Object) = Each accessed

session/number of session in weblog×100 (1)



Equations f or Computing the Eff iciency



Equations f or Computing the Eff iciency



Results of International Islamic University Web data



Results of Hazara University web server data set



Web log Data Set of National University of

Science and Technology



Overall Generalize Comparison of Results



Conclusions

The results proved that proposed model

technique is efficient in web page prediction

by predicting the resources local to the user

Three data sets of web servers are tested on

existing model and proposed technique

We simulated the client server environment

for improving more efficient results as

compared.



REFERENCES

[1] Greg Barish, Katia Obraczka, WorldWideWeb Caching: Trends and

Techniques, IEEE Communications Magazine, Inf ormation Science

Institute University of Southern Calif ornia, Los Angeles, USA, pp: 178-

184, May 2000.

[2] O Kit Hong, Fiona Robert, P.Biuk Aghiai, AWeb Prefetching Model BasedContent Analysis.

http://www.sf tw.umac.mo/~robertb/publications/MITC99/MITC99.pdf,

accessed on July 10, 2009.

[3] Josep Dom`enech, Julio Sahuquillo, Jos´e Ana Gil, Ana Pont, The Impact

of theW

eb Prefetching Architecture on the Limits of Reducing UsersPerceived Latency, In proceeding WI '06 of IEEE/WIC/ACM International

Conference onWeb Intelligence, Hong Kong, pp: 740744, 2006.

[4] B.delaOssa, J.A. Gil, J.Sahuquillo, A.Pont, Web Prefetch Perf ormance

Evaluation in a Real Environment, IFIP/ACM Latin America Networking

Conference, pp: 8, 11 October 2007.



REFERENCES

[5] B.de la Ossa, A. Pont, J.Sahuquillo, J.A. Gil, Referrer Graph: a low cost web

prediction algorithm, Proceedings of the 25th ACM Symposium on Applied

Computing (ACM SAC 2010), Switzerland, pp: 831-838, 2010.

[6] George Pallis, Athena Vakali, Jaroslav Pokorny, A clustering-based prefetching

scheme on aWeb cache environment, Journal of Computers and Electrical

Engineering, ACM, Vol.34, Issue: 4, pp:309-323, 2008.

[7] Christos Bouras, Agisilos Konidaris, A Most Popular Approach of Predictive

Prefetching on aWAN to Eff iciently ImproveWWW Response Times, In

proceedings of Springer, Berlin, pp: 344-351,2004.

[8] Bamshad Mobasher, Honghua Dai,Tao Luo, Miki Nakagawa, Using Sequential

and Non- Sequential Patterns in PredictiveWeb Usage Mining Tasks, In

proceedings IEEE of International Conference of Data Mining(ICDM), USA, pp:

669-672, 2002.

[9] Beihong Jin, Sihua Tian, Chen Lin, Xin Ren, Yu Huang, An Integrated Prefetching

and Caching Scheme f or MobileWeb Caching System, In proceeding of Sof tware

Engineering, Artif icial Intelligence/Networking, and Parallel/Distributed

Computing, Q ingdao, China, pp: 522 -527, 2007.



REFERENCES

[10] Jaideep Srivastava, Robert Cooleyz, Mukund Deshpande, Pang-Ning Tan, Web

Usage Mining: Discovery and Applications of Usage Patterns fromWeb Data,

ACM SIGKDD, New York, USA, Vol: 1, Issue: 2, January 2000.

[11] Q iang Yang, Zhen Zhang, Model based Predictive Prefetching, In proceeding of

International workshop of Database and Expert System Application, IEEE

Computer Society,Washington, USA, pp: 291-295, ISBN: 0-7695-1230-5, 2000.[12] Naveed Ahmad, Owais Malik, Mahmood ul Hassan, Muhammad Shuaib Qureshi,

Asim Munir, Reducing User Latency inWeb Prefetching Using Integrated

Techniques, In proceeding of International Conference on Computer Network and

Inf ormation Technology (ICCNIT), pp: 175-178, 11-13th July 2011.

[13] Q iang Yang, Haining Henry Zhang, Tianyi Li, Mining Web Logs f or Prediction

Models inWWW Caching and Prefetching, International Conference onKnowledge Discovery and Data mining in proceedings of the seventh ACM

SIGKDD, San Francisco, Calif ornia, USA, pp:473-478, 2001.

[14] Miguel Gomes da Costa Júnior, Zhiguo Gong, Web Structure Mining:An

Introduction, Proceedings of the IEEE International Conferenceon Inf ormation

Acquisition, Hong Kong and Macau, China, June 2005.



Thanks for your attention

Optimizing Predictive Pre Fetching in Multi-Client

Documents