WEB WORKLOAD GENERATORS

A Survey Focusing on User Dynamism Representation

ul Pe

na-Ortiz

Computer Engineering Department (DISCA), Polytechnic University of Valencia, Cam

ı de Vera s/n, Valencia, Spain

Julio Sahuquillo, Jos

e A. Gil, Ana Pont

Computer Engineering Department (DISCA), Polytechnic University of Valencia, Cam

ı de Vera s/n, Valencia, Spain

Keywords:

Web performance evaluation, Dynamic users’ navigation, Dynamic contents, Modeling dynamic workload.

Abstract:

The evolution of the World Wide Web from hypermedia information repositories to hypermedia distributed

applications and services oriented architectures (SOA) has introduced new features in the current and incoming

web. An important feature is the dynamism of its contents and services, which induces the dynamism of

the client behavior. This feature represents a major constrain when modeling and generating current web

workload.

In this paper, we ﬁrst review the state of the art for web workload generation, focusing on the approach based

on workload models. After that, we analyze a representative subset of the state of the art workload generators

that use workload models, concentrating on those model characteristics that represent dynamism in the work-

load generation. The study reveals that ﬁve generators present some capabilities to reproduce this dynamism,

but only the GUERNICA approach improves the dynamic workload generation by using users’ behavior mod-

els. Finally, we discuss GUERNICA and describe how it generates dynamic workload by addressing both user

behavior and workload distribution.

1 INTRODUCTION

As a consequence of the recent and incessant changes

in web technology, new types of hypermedia dis-

tributed applications, hereafter web applications, and

distributed services have become more frequent and

familiar to web users. For instance, e-commerce,

on-line booking systems, on-line auction systems,

Google Mashups and Amazon Web Services are only

few examples that manifest how web sites are evolv-

ing from hypermedia information repositories to hy-

permedia distributed applications and distributed ser-

vices. Due to the increase and popularity of these ap-

plications and services, typical of the Web 2.0, dy-

namic contents and services are becoming more and

more frequent, suggesting the review of widely ac-

cepted paradigms and models. The Web 2.0 technol-

ogy involves important changes in the web site ar-

chitecture (Treese, 2006), for instance, AJAX, ﬂash,

tagging, social network, or services oriented architec-

tures (SOA) (Erl, 2005). These changes are more rel-

evant and meaningful in the recently incoming Web,

also refereed to as Future Internet (Tselentis, 2009).

This evolution also implies signiﬁcant changes in web

users’ behavior (O’Reilly, 2005) because new appli-

cations promote a dynamic navigation, providing an

experience much closer to desktop applications than

the traditional static web pages, e.g., syndication of

website content or web information systems (Houben

et al., 2004).

As any system that is continuously changing, both

in the offered applications and in its infrastructure,

performance studies are important to evaluate novel

proposals when designing new web-related systems,

such as services oriented architectures, web servers,

or proxy policies. These studies require from the use

of an accurate and representative workload models in

order to guarantee the representativeness and valid-

ness of the results.

In the case of the Web, an appropriate web work-

load model is a major concern in order to reproduce

client workloads. These models can be used to eval-

uate the system performance and to test applications

and services since they reproduce the behavior of their

119

Peña-Ortiz R., Sahuquillo J., A. Gil J. and Pont A..

WEB WORKLOAD GENERATORS - A Survey Focusing on User Dynamism Representation.

DOI: 10.5220/0003274801190126

In Proceedings of the 7th International Conference on Web Information Systems and Technologies (WEBIST-2011), pages 119-126

ISBN: 978-989-8425-51-5

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

potential clients. In this context, a web workload

model should reproduce the HTTP requests that a real

web server typically receives in daily working ses-

sions.

Nevertheless, the implicit dynamism in users’ be-

havior and in the generation of web content makes

hard to design accurate web workload models that

represent users’ navigations. On the one hand, users

require customized information and advanced ser-

vices from web applications. This requirement im-

plies that the contents are dynamically generated from

different data sources. On the other hand, along a

typical navigation, users use to select pages to visit

depending on the previously visited ones, and particu-

larly on their contents or characteristics such as the re-

sponse time. Thus, the dynamism in the content gen-

eration implies that the users’ navigation may differ

depending on the contents generated by the web ap-

plication during a given period of time. In summary,

the dynamic content introduces dynamism in users’

navigation, that is, it produces dynamic users’ behav-

ior. For example, a high percentage of users’ naviga-

tion sessions begins searching a dynamic resource in

a specialized site and then it continues visiting one or

more sites, depending on the results of the previous

searches.

In general, the main challenges when modeling

web workloads are: i) how to model users’ dynamism,

ii) how to represent the different roles that users play

in the web, and iii) how to model continuous changes

in users’ behavior (Weinreich et al., 2006).

In a previous work (Pe

na-Ortiz et al., 2005) we

proposed the GUERNICA approach to model the dy-

namism of the web workload based on users’ behav-

ior models. This paper extends that work in sev-

eral ways. First, we review the state of art on web

workload generators, software products that are de-

signed and implemented to generate web workload,

and classify them in three main groups according to

their capability to generate web workload and/or to

model the users’ dynamic behavior. Second, we com-

pare GUERNICA to other web workload generators.

The study reveals that ﬁve web workload generators

present some capabilities to reproduce this dynamism,

but only our approach improves the dynamic work-

load generation by using users’ behavior models.

The remaining of this paper is organized as fol-

lows. Section 2 reviews, analyzes and classiﬁes a rep-

resentative subset of workload generators. Section 3

summarizes highlights some GUERNICA character-

istics. Finally, Section 4 presents some concluding

remarks.

2 WEB WORKLOAD

GENERATORS

Dynamic web applications and services have induced

continuous changes in users’ navigation patterns, thus

making it difﬁcult to characterize the web workload.

Currently, we can reproduce the client behavior by

using either traces or workload models. Traces log the

sequence of HTTP requests and commands received

by a web application or a web server during a certain

period of time and under speciﬁc conditions. There-

fore, traces are obtained in a particular environment

(e.g., server process speed or network bandwidth) for

a speciﬁc application. This means that if any system

parameter changes (e.g., new contents in a news por-

tal), the resulting trace might differ. Therefore, the

main concern in trace-based study is the representa-

tiveness potentially achieved, especially when the re-

quests received by a given web server exhibit a high

variability because they only reproduce a subset of

their clients. Consequently, these models are not ap-

propriate to model changes in the client behavior.

On the other hand, parameterizable workload

models are abstractions of the real workload that

hide those characteristics not relevant for a particular

study. Unlike the previous models, this kind of mod-

els is accurate enough when reproducing the client

behavior or when evaluating the performance of web

applications. Parameterizable workload models gen-

erate sequences of HTTP requests similar to the real

sequences and different scenarios can be conﬁgured

by properly setting the corresponding parameters.

The representation of the dynamic users’ behavior

has been addressed in some web workload characteri-

zation studies when deﬁning dynamic web site bench-

marks (Amza et al., 2002) or when studying the per-

formance and scalability of the technology used to de-

velop dynamic sites (Cecchet et al., 2002). However,

user dynamism is complex by nature, and current re-

sults are still far from being precise and satisfactory.

Therefore, more research efforts must be done in this

direction in order to provide more accurate tools to

model and evaluate web performance. In this sense,

the Customer Behavior Model Graph (Menasc

e and

Almeida, 2000) was introduced to be used as input to

a workload model of an e-business site.

Workload models are the basics of workload gen-

erators. They are software products designed and

implemented to generate HTTP requests. They are

ﬂexible tools useful to address tuning or capacity

planning studies but, unfortunately, current genera-

tors only represent the users’ dynamic behavior in a

partial way; for instance, they cannot represent user

navigation changes as a response to the quality of the

content, the quality of its generation, or the character-

WEBIST 2011 - 7th International Conference on Web Information Systems and Technologies

120

istics of the last visited pages.

Comparing workload generators is a laborious and

difﬁcult task since they offer a large amount and diver-

sity of features. In this paper we compare generators

according to a wide set of features and capabilities.

Below, the twelve features and capabilities used are

deﬁned to ease the understanding of the comparison

study.

• Distributed Architecture. It refers to the ca-

pability to distribute the processes of generation

among the different nodes. The distribution of

the workload generation signiﬁcantly helps to im-

prove the workload accuracy.

• Mathematical-based Architecture. This fea-

ture represents the capability to use mathemati-

cal models to deﬁne the workload. These models

allow to improve the workload quality by using

them as workload parameters (e.g., users’ behav-

ior models or simulation architectures.)

• Business-based Architecture. This refers to

the capability to model web application busi-

ness logic. When deﬁning a testing environment,

the architecture simulator should implement the

same features as the real environment (e.g., e-

commerce architectures typically include a cata-

log, a product searcher, or a payment gateway).

• Client Parametrization. This is the ability to

parameterize generators nodes (e.g., number of

users, allowed navigations set, or changes be-

tween navigations). In general, web dynamism

highlights the need for a workload characteriza-

tion based on parameters, and specially the related

user behavior.

• Workload Types. Some generators organize the

workload in categories or types, each one mod-

eling a given user proﬁle (e.g., searcher or buyer

user proﬁles).

• Modeling of the Web Application Business

Logic (Business Test). This capability permits to

deﬁne functional tests related to a real web appli-

cation. These tests allow to guarantee the appli-

cation correctness (i.e., the application provides

the deﬁned functionality, which fulﬁlls the qual-

ity and assurance requirements).

• Multiplatform. Is referred to a computer soft-

ware that are implemented and inter-operate on

multiple computer platforms.

• Differences between LAN and WAN. Simula-

tions usually run in LAN environments, and most

of the current simulators cannot model differences

between LAN and WAN, where applications are

usually located.

• Ease of Use. The generator should be a

friendly application carrying out usability guide-

lines, mainly in commercial products.

• Performance Reports. The results elaborated by

the generation process are usually presented by

using graphical plots.

• Open Source. This feature allows the develop-

ers’ community to develop extensions or different

generation alternatives over the generator archi-

tecture.

• Dynamic Workload. This is the main feature we

are interested in, because the dynamism in con-

tents and users is the most relevant characteristic

in the current web.

Table 1 summarizes the studied software packages

used to generate web workload as well as the grade

(full or partial) in which they fulﬁll the features de-

scribed above. By nature, this software can be classi-

ﬁed in benchmarks and generators. In this paper, the

latter group is broken down according to the genera-

tor offers capabilities to model the dynamic workload

or not. In other words, as observed in Table 1, three

main groups have been identiﬁed as discussed below.

• Group I: Benchmarks that model the client and

server paradigm in web context, but they do not

consider dynamism in web workload generation.

Webstone was designed by Silicon Graphics in

1996 and it is currently commercialized as a

Mindcraft product. It bases its operation on units

of execution located on the client. These units

simulate the clients’ behavior when accessing web

resources. These units can run on several ma-

chines in order to generate requests to the server

resources, which can be classiﬁed in different cat-

egories according to their size.

SPECweb99 was designed to measure the perfor-

mance of systems offering services in the web.

This generator, commercialized by SPEC, gener-

ates both static and dynamic requests. The work-

load is distributed among several clients, which

are managed by a central client that collects data

from the other ones. The clients generate requests

according to certain categories that are deﬁned af-

ter studying different web applications. In addi-

tion, clients verify the result of each request. This

generator evaluates the architectural performance

in order to provide a generic proﬁle of web appli-

cations.

SURGE was developed by Barford and Crovella

(Barford and Crovella, 1997) with the goal of

measuring the server behavior while varying the

user load. The generator performs an analytical

WEB WORKLOAD GENERATORS - A Survey Focusing on User Dynamism Representation

121

Table 1: Workload generators and grade in which features are fulﬁlled.

Feat./Cap.

Gen.

Webstone

SPECweb99

SURGE

Polygraph

S-clients

Webjamma

Deluge

Hammer

PTester

Siege

Httperf

Autobench

TPC-W

Webload

Mercury

JMeter

GUERNICA

Mathematical-based arch.     N

Dsitributed architecture      

Business-based arch. N   N N N N

Client parametrization   N  N     

Workload types  

Business test  N

LAN and WAN  N

Multiplatform        N  N       

Ease of use    

Performance reports            

Open source            N

Dynamic workload N N N N 

Group I Group II Group III

 Full support N Partial support

characterization of the user load and a set of math-

ematical models that generate the HTTP requests

in the server.

Web Polygraph was developed at the California

University (Rousskov et al., 1999) and it is a

workload generator based on mathematical mod-

els which simulates both the client and the server.

It also allows to introduce real components in-

stead of mathematical models.

• Group II: Workload generators that reproduce

web clients behavior without considering dy-

namism.

S-Clients was developed by Banga and Druschel

(Banga and Druschel, 1999). It generates load

sporadic tips to improve the beneﬁts provided by

the server. S-Clients splits the process of gener-

ating traced HTTP requests in two subprocesses:

one for obtaining the connection and the other for

recovering the content, so enabling a relative par-

allelism.

Webjamma was developed by Virginia Tech’s Net-

work Research Group. It is aimed at serving as a

baseline for developing a future generator. This

generator works in a simple way by taking a URL

ﬁle that provides the source of the HTTP requests

to be generated, so it cannot represent dynamic

users’ behavior. To this end, it uses a multipro-

cessing architecture based on generation nodes.

Deluge was developed by Thrown Clear Produc-

tions. It is a basic and open source solution that

includes three main components: i) dlg proxy

records HTTP requests, ii) dlg attack gener-

ates workload by reproducing recorded users’ re-

quests, so it does not allow dynamic HTTP re-

quests generation, and iii) dlg eval elaborates

statistics from the generated results.

Scripts and tools, which are usually basic and

open source approaches, only allow to capture

HTTP requests and to reproduce them for stress-

ing purposes; e.g., HAMMERHEAD, PTESTER,

SIEGE, HTTPERF or AUTOBENCH.

• Group III: Generators that model web workload

considering the dynamism.

TPC-W (Smith, 2000) was deﬁned by TPC. It

is oriented to e-commerce web transactions, pro-

viding both models of business-client (B2C) and

business-business (B2B). It is aimed at evaluating

performance of architecture on a generic proﬁle

of web applications. It examines real features of

e-commerce applications: security, catalog, etc.

The application proﬁle is conﬁgured by choosing

the desired features, so allowing to represent dif-

ferent user behaviors. TPC-W partially models

dynamic users’ behavior by means of these users’

proﬁles.

Webload (RadView Software, 2003) is web work-

load generator commercialized by RadView. It

is oriented to explore the performance of criti-

cal applications, quantifying resource usage in the

server. Webload is aimed at evaluating the cor-

rectness of a given application (testing). This gen-

erator includes two main modules: the ﬁrst mod-

ule is on the client side and enables it to execute

request agendas (navigation scripts) on the server

side; the second module is on the server side and

estimates performance gains. Webload presents

WEBIST 2011 - 7th International Conference on Web Information Systems and Technologies

122

some capability to model users’ behavior by us-

ing agendas.

MERCURY LoadRunner is a software, commer-

cialized by Mercury, for functional and perfor-

mance web testing. It generates the speciﬁc work-

load of a web application by simulating the con-

current navigations of a typical users set. The gen-

erator supports the recording of the navigation and

the deﬁnition of users’ families to generate web

workload. The software also includes a versatile

module that permits the graphic representation of

the representation. LoadRunner allows to deﬁne

users’ navigations and partially dynamic users’

behavior by using Javascript language to deﬁne

these navigations.

JMeter is a solution for generating workload and

evaluating web server performance presented by

Apache Project. It is written entirely in Java and

provides an easily conﬁgurable and visual API,

evaluating the performance of client side web ap-

plications. It presents partial capability to model

dynamic workload by deﬁning a navigation test

based on patterns (e.g., regular expressions).

GUERNICA is a web workload generator de-

signed to generate dynamic workload by model-

ing dynamic users’ behavior. In addition, its de-

sign and implementation is intended to achieve a

solid baseline generator as a ﬁrst step to support

future improvements and extensions with the aim

of becoming a commercial product.

Most of the presented generators do not model dy-

namic workload because: i) they are simulation ap-

proaches that do not reproduce real workload (e.g.,

Webstone or Web Polygraph), ii) they are based on

HTTP traces (e.g.,Webjama, S-Clients or the scripts

and tools solutions) or iii) they are based on workload

models that do not consider dynamism as a parameter

(e.g., SPECWeb99 or SURGE).

In summary, we can conclude that only ﬁve of the

studied approaches (i.e., Webload, TPC-W, JMeter,

Mercury and GUERNICA) present some capability

to model users’ dynamic behavior and to generate dy-

namic workload, but only GUERNICA presents fully

capability. This generator totally supports seven of

the twelve described features and partially supports

the remaining ones.

3 GUERNICA

GUERNICA (Universal Generator of Dynamic Work-

load under Web Platforms) is a workload generator

based on the Dweb model (Pe

na-Ortiz et al., 2009),

a recent workload model able to characterize the web

dynamism in an accurate way. Three main concepts

allow Dweb to deal with dynamic workload: naviga-

tion, workload test, and workload distribution. The

two former permit to characterize the workload dy-

namism by means of web users’ behavior models.

The latter concept allows to improve the workload ac-

curacy by distributing its generation among different

machines.

Its design is the result of the cooperation among

the Web Performance Research Group (Polytechnic

University of Valencia), Intelligent Software Compo-

nents and the Instituto Tecnol

ogico de Inform

atica;

thereby, bridging the gap between academia and in-

dustry.

GUERNICA applications and Dweb model con-

cepts permit to carry out the workload test process by

considering different levels of dynamism in users’ be-

havior when characterizing web workload. This pro-

cess consists of four main phases (see Figure 1) de-

tailed below.

1. Navigation Deﬁnition. This phase deﬁnes the

ﬁrst level of dynamism by modeling the users’ dy-

namism when interacting with the dynamic web

contents (typical of dynamic web). For instance,

when a user runs a web searcher to make a query,

he or she usually continues visiting one or more

sites determined by the search outcome. If the

query is successfully resolved (i.e., a list of re-

sults is obtained), it is likely that the user will ei-

ther i) visit the ﬁrst site of the list, or ii) reﬁne the

search. On the other hand, if the response time is

too longer, the query might be canceled. In other

words, each user request depends not only on the

response itself but also on other issues related to

the quality of service (e.g., response time length

or content amount).

Owing to dynamic web contents, the users’ dy-

namism is modeled by using the navigation con-

cept. Figure 2 shows the navigation tree corre-

sponding to a Google search navigation.

Two main parts can be distinguished in the navi-

gation:

(1) The upper part of the diagram (before reaching

branch b1) shows the two ways in which the

search can be initiated:

a. On the left side, the user makes use of a search

toolbar of a web browser (e.g., Google tool-

bar for Mozilla Firefox) to make the query di-

rectly.

b. On the right side of the ﬁgure, the user

reaches branch b1 after the Google.HOME

node, where the user requests the main page

(http://www.google.com) to web searcher en-

gine. After that, she or he waits for a while

WEB WORKLOAD GENERATORS - A Survey Focusing on User Dynamism Representation

123

Figure 1: Testing phases in GUERNICA.

(time referred to as the user thinking time) and

then the user makes the query.

(2) In the below side of the diagram, the user ana-

lyzes the results:

a. If the web search engine provides results (path

from b1 to b2) the user analyzes them. After

that, she or he can reﬁne the results by making

a new query, reﬁned query in the ﬁgure (path

from b2 to b1), or access the top n sites pro-

vided and ﬁnish the navigation (path from b2

to black dot through X TH RESULTS.HOME

node).

b. On the contrary, that is, if no results are pro-

vided (path from b1 to b3), the user can make

a new query, other query in the ﬁgure (path

from b3 to b1), or ﬁnish the process (path from

b3 to black dot).

CARENA (Ni

no et al., 2005) is a Mozilla plu-

gin that helps GUERNICA to deﬁne users’ nav-

igations. It captures real users’ navigations that

GUERNICA converts to its internal representa-

tion.

2. Performance Test Deﬁnition and Execution

Conﬁguration. It deﬁnes the workload of the tar-

get site by using the workload test and workload

distribution concepts.

Workload test concept allows to deﬁne the second

level of dynamism. This level is related to the

roles that users play in the web and the continuous

changes of these roles deﬁning users’ behaviors.

For example, assume people at the Import and Ex-

port Department in a typical multinational com-

pany that use to access the web during their job.

Assume also that the company has got a web

ERP

(intranet) which allows web access. Most

of the time, workers use Internet for professional

purposes (e.g., intranet navigations, suppliers sites

navigations, or professional web searches), but

sometimes they use the web for leisure purposes

(e.g., reading the news with Google Reader, per-

forming personal searches, or checking the mail).

So we can distinguish between two roles when a

department member navigates the web: working

behavior (professional navigations), and leisure

behavior (personal navigations).

The Dweb model uses the workload test concept

to model user behaviors and changes between be-

haviors. Figure 3 deﬁnes the working and leisure

behaviors of the example, and the likelihood to

change between behaviors by using balanced arcs

(the arc weight is the probability to change from

the source behavior to the destination behavior).

These behaviors are deﬁned as automatons, where

their nodes represent navigations, and their bal-

anced arcs indicate the transitions between nav-

igations (the arc weight indicates the probability

to take that arc).

Finally, Workload distribution concept allows to

Enterprise Resource Planning (ERP) is an application

used by large organizations to manage inventory, resources,

and business processes across departments.

WEBIST 2011 - 7th International Conference on Web Information Systems and Technologies

124

Figure 2: Google Search navigation pattern.

improve the workload accuracy by distributing its

generation among different machines.

3. Tests Execution. This phase executes workload

tests gathering performance statistics. The plan-

ner application controls and plans the workload

test. The probe client application is aimed at eval-

uating from the point of view of a typical user

(e.g., response time or application correctness) the

major functionalities of a given web application,

while it is stressed by the workload generators.

Results collected by generators and probe clients

are given to the planner, which groups, classiﬁes

and reaches a consensus among them in order to

obtain an uniform set of results.

4. Results Analysis. Finally, GUERNICA analyzes

the performance of the system under test and rep-

resents the performance indexes by using graphi-

cal plots. Regarding 3D representation, it repre-

sents the results in an external format based on

an ontology that SemViz (Bl

azquez et al., 2008)

(implemented as an applet) can read and display

graphically. SemViz visualizes knowledge based

on ontologies by using 3D technology.

4 CONCLUSIONS

Due to the increase in the number and popularity of

web applications and distributed services such as e-

commerce, social networks or Amazon Web Services,

dynamic contents are becoming more and more fre-

quent. For example, Web 2.0 applications mainly fo-

cus on users who are the potential clients by provid-

ing them an experience closer to desktop applications.

One of the most important mechanisms to achieve this

goal is the dynamism present in the web content (e.g.,

personalized contents or publicity), which induces the

dynamism of users’ behavior in the current web. This

feature is the main shortcoming when modeling and

generating real web workload.

In this paper we have analyzed the characteristics

of a representative subset of the state of the art work-

load generators, focusing on the capability to repre-

sent user dynamism in the current and incoming web.

We found ﬁve workload generators that present some

capability to represent workload dynamism although

only GUERNICA allows to fully represent this dy-

namism. With the aim of illustrating how this fact has

been achieved, we described the workload test pro-

cess used in the generator side to address both user

WEB WORKLOAD GENERATORS - A Survey Focusing on User Dynamism Representation

125

Figure 3: Working and leisure behaviors workload test.

dynamic behavior and workload distribution.

ACKNOWLEDGEMENTS

This work has been partially supported by the Spanish

Government Grant (CICYT TIC2001-1374-C03-02

and FIT-340000-2004-236), Spanish Ministry of Edu-

cation and Science and the European Investment Fund

for Regional Development Grant (TSI 2005-07876-

C03-01) and IMPIVA Grant (IMIDTD/2004/92,

IMIDTD/2005/15).

REFERENCES

Amza et al. (2002). Speciﬁcation and Implementation of

Dynamic Web Site Benchmarks. In 5th Workshop on

Workload Characterization.

Banga, G. and Druschel, P. (1999) Measuring the capacity

of a Web server under realistic loads. In 8th WWW,

2(1-2):69–83.

Barford, P. and Crovella, M. (1997). An Architecture for

a WWW Workload Generator. In WWW Consortium

Workshop on Workload Characterization.

azquez et al. (2008). Trends on Legal Knowledge, the

Semantic Web and the Regulation of Electronic Social

Systems, Chapter Visualization of Semantic Content.

Cecchet et al. (2002). Performance and scalability of EJB

applications. SIGPLAN Not., 37(11):246–261.

Erl, T. (2005). Service-Oriented Architecture: Concepts,

Technology, and Design. Prentice Hall PTR.

Houben et al. (2004). Modeling User Input and Hypermedia

Dynamics in Hera. In Web Engineering - 4th Interna-

tional Conference, Vol. 3140, 60–73.

Menasc

e, D. A. and Almeida, V. A. F. (2000). Scaling for

E-Business: Technologies, Models, Performance, and

Capacity Planning. Prentice Hall PTR.

no et al. (2005). Carena: A tool to capture and replay

web navigation sessions. Third IEEE/IFIP Workshop

on End-to-End Monitoring Techniques and Services.

O’Reilly, T. (2005). What is Web 2.0. Design Patterns and

Business Models for the Next Generation of Software.

O’Reilly Online Publishing.

na-Ortiz et al. (2005). Modeling continuos changes of

the user’s web dynamic behavior in the WWW. In

Fifth International Workshop on Software and Perfor-

mance, 175–180.

na-Ortiz et al. (2009). Dweb model: Representing

Web 2.0 dynamism. Computer Communications,

32(6):1118–1128.

Rousskov et al. (1999). The ﬁrst ircache web cache bake-

off. Technical report, National Laboratory for Applied

Network Research.

Smith, W. D. (2000). TPC-W: Benchmarking An Ecom-

merce Solution. Technical report, Intel Corporation.

Technical report, RadView Software. (2003). WebLOAD

6.0 The WebLOAD difference.

Treese, W. (2006). Web 2.0: Is It Really Different ? net-

Worker, 10(2):15–17.

Tselentis, G. (2009). Towards the Future Internet: A Euro-

pean Research Perspective. IOS Press Inc.

Weinreich et al. (2006). Off the beaten tracks: exploring

three aspects of web navigation. In 15th WWW, 133–

142.

WEBIST 2011 - 7th International Conference on Web Information Systems and Technologies

126