MONITORING SERVICE COMPOSITIONS IN MoDe4SLA

Design of Validation

Lianne Bodenstaff

∗

, Andreas Wombacher, Roel Wieringa

University of Twente, Enschede, The Netherlands

Michael C. Jaeger

Berlin University of Technology, Germany

Manfred Reichert

University of Ulm, Ulm, Germany

Keywords:

Service composition, Monitoring, Management, Validation.

Abstract:

In previous research we introduced the MoDe4SLA approach for monitoring service compositions.

MoDe4SLA identiﬁes complex dependencies between Service Level Agreements (SLAs) in a service com-

position. By explicating these dependencies, causes of SLA violations of a service might be explained by

malfunctioning of the services it depends on. MoDe4SLA assists managers in identifying such causes. In

this paper we discuss how to evaluate our approach concerning usefulness for the user as well as effectiveness

for the business. Usefulness is evaluated by experts who are asked to manage simulated runs of service com-

positions using MoDe4SLA. Their opinion on the approach is an indicator for its usefulness. Effectiveness

is evaluated by comparing runtime results of SLA management using MoDe4SLA with runtime results of

unsupported management. Criteria for effectiveness are cost reduction and increase in customer satisfaction.

1 INTRODUCTION

Service monitoring and management are challenging

research areas. On the one hand research exists on

composing services such that an optimal service level

can be offered to the customer (Zeng et al., 2004),

(Cardoso et al., 2004). On the other hand, research

exists on monitoring performance of each service in

the composition (Tosic et al., 2005). However, to our

knowledge there is no research on monitoring compo-

sitions which take dependencies between correspond-

ing services into account. Consequently, it is difﬁcult

to determine why Service Level Agreements (SLAs)

of composite services are violated. Often badly per-

forming services on which the composition depends

cause these violations. Managers face the challenge

of identifying these problematic services. Therefore,

∗

This research has been supported by the Dutch Organi-

zation for Scientiﬁc Research (NWO) under contract num-

ber 612.063.409

support in this tedious task is much needed. Effec-

tively managing compositions results in competitive

service level offerings to customers with maximum

proﬁt for the business. Furthermore, insights on ser-

vices run by other providers, helps managers to plan

negotiation strategies concerning SLAs. With the

MoDe4SLA approach we monitor dependencies be-

tween services and their impact on the composition

(Bodenstaff et al., 2008). However, so far, we have

not evaluated MoDe4SLA. In this paper, we describe

both our evaluation approach and evaluation criteria.

We aim at evaluating two aspects of MoDe4SLA.

On the one hand we plan to evaluate how useful the

approach is for managers burdened with actual main-

tenance of compositions. On the other hand we want

to evaluate how effective MoDe4SLA is when main-

tenance is done using this approach. Since we do

not have access to real composition monitoring data,

we implement a simulator for running composite ser-

vices. Experts from both industry and academia are

asked to manage these generated compositions both

114

Bodenstaff L., Wombacher A., Wieringa R., C. Jaeger M. and Reichert M. (2009).

MONITORING SERVICE COMPOSITIONS IN MoDe4SLA - Design of Validation.

In Proceedings of the 11th International Conference on Enterprise Information Systems - Software Agents and Internet Computing, pages 113-121

DOI: 10.5220/0002003401130121

 SciTePress

Composition Manager

SLA Analysis

<party> ...

</party>

<offer> ...

</offer>

</guarantee>

<party> ...

</party>

<offer> ...

</offer>

</guarantee>

Dependency Analysis

State 1

State A State B

State 2

Impact Factor Analysis

Impact(1) = 5

Impact(2) = 3

Impact(A) = 5

Impact(B) = 3

Presentation

State 1

State A State B

State 2

Log Extraction

Application

Log

20081204-14:45

INFO ...

20081204-14:45

INFO ...

20081204-14:45

WARN ...

20081204-14:45

DEBUG ...

20081204-14:45

INFO ...

20081204-14:45

INFO ...

20081204-14:45

INFO ...

20081204-14:45

WARN ...

20081204-14:45

DEBUG ...

20081204-14:45

INFO ...

20081204-14:45

INFO ...

20081204-14:45

INFO ...

20081204-14:45

WARN ...

20081204-14:45

DEBUG ...

20081204-14:45

INFO ...

1 2

Figure 1: MoDe4SLA Approach.

with and without using MoDe4SLA.

To evaluate usefulness, we plan to interview ex-

perts, asking them to make a statement on how useful

they perceivedthe approach when managing the com-

positions. To evaluate effectiveness we plan to test

performance of compositions managed by experts us-

ing MoDe4SLA and of compositions managed by the

control group not using MoDe4SLA. The level of cus-

tomer satisfaction and proﬁts made by the company

are key criteria for evaluating effectiveness.

In Section 2 we discuss background needed for the

understanding of the paper. We elaborate on the eval-

uation approach in Section 3, and conclude with re-

lated work and an outlook in Sections 4 and 5.

2 BACKGROUND: MoDe4SLA

Companies offering composite services to their cus-

tomers struggle to manage these complex constella-

tions. Different services are provided with different

levels of quality. These services stem from differ-

ent providers, and have different levels of impact on

the composition. Consider, for example, a service

provider which allows ﬁnancial institutes to check

creditworthiness of potential customers. This service

is composed of services querying databases, and a

payment service where several payment options (e.g.,

PayPall and credit card) are provided. The company

has an SLA with each of these service providers. The

composite service offered to the ﬁnancial institutes is

dependent on all these services (e.g., payment ser-

vices and database querying services). To meet the

SLA with its customers the company faces the chal-

lenge of properly managing its underlying services,

particularly since SLAs are often violated. For each

of these violations the company has to determine how

big their impact is on the composition, and it has

to decide on which SLAs to renegotiate. Generally,

complexity of this decision process grows with the

number of services being involvedin the composition.

Fig. 1 depicts the implementation of MoDe4SLA:

• At design time, we analyze the relations between

different services and the composition with re-

spect to the agreed response time and costs of the

different providers (step 1 in Fig. 1).

• The result of this dependency analysis is the input

for a subsequent impact analysis (step 2 in Fig. 1).

The latter results in an impact model with intuitive

graphical representation.

• During runtime, event logs are analyzed using the

event log model, ﬁltering events referring to the

services, and their SLA statements for the com-

position (step 3 in Fig. 1). These results, together

with the impact analysis and dependencies, con-

stitute the input for the monitoring interface (step

4 in Fig. 1). The latter enables time efﬁcient

management and maintenance of the composition

(step 5 in Fig. 1).

While we have described the MoDe4SLA frame-

work in (Bodenstaffet al., 2008), this section provides

a summarized description. Although MoDe4SLA is

especially helpful in complex scenarios, we illustrate

it by means of a simpliﬁed example. In this exam-

ple a company offers composite service NewsRequest

MONITORING SERVICE COMPOSITIONS IN MoDe4SLA - Design of Validation

115

IF: 0.5 IF: 0.375 IF: 0.45 IF: 0.2

0.75 0.25

Figure 2: Impact Models.

IF: 0.45 IF: 0.43 IF: 0.48 IF: 0.2

Figure 3: Feedback Model.

containing up-to-date news to its customers on a spe-

ciﬁc search query. Customers pay per invocation of

the service. To gather the necessary information, the

company invokes two content providers and sends in-

formation of the fastest responding one to the cus-

tomer (cf. Fig. 4). Content provider 1 (CP1) is its

main provider, offering fast response times (in 99.9%

of the cases within 3 ms), accurate information, but

for high costs (0.20 cents per invocation). The sec-

ond content provider (CP2) constitutes the backup of

the company in case the ﬁrst one fails. This provider

offers a slower response time and guarantees less ac-

curate information but for a lower price.

Quality of service the company can offer to its

customers is determined by the services it depends on.

In our example (cf. Fig. 4) each service has an objec-

tive on response time as well as on cost. An impact

model is derived for both objectives (cf. Fig. 2) which

depicts dependencies between the different services

for one run of the composition. Therefore, the im-

pact models of the two objectives are different. E.g.,

when invoking both services in NewsRequest while

using only data from the fastest responding one, re-

sponse time depends solely on this service, while cost

depends on both services. In addition an expected im-

pact factor indicating the impact a service has on the

composition is calculated for each service, per objec-

tive. E.g., CP1 in the time impact model is expected

to be chosen 75% of the time (cf. Fig. 2), with an av-

erage response time of 3ms (cf. Fig. 4). The expected

contribution per composition invocation for CP1 is

therefore 3ms times 0.75 divided by 5ms, which is an

SLA with CP1:

-Cost: 0.20 euro per invocation.

-Time: 99.9% within 3 ms.

SLA with CP2:

-Cost: 0.15 euro per invocation

-Time: 99% within 4 ms.

SLA with Customer:

-Cost: invocation first 100x

0.50 euro, thereafter 0.40 euro.

-Time: 99% within 5 ms.

Figure 4: Service Composition: News Request.

impact factor of 0.45. Intuitively, services with a high

impact factor require more attention when managing

the composition than services with low impact.

At runtime the impact models are colored based

on monitoring results (cf. Fig. 3) in comparison to

the estimations made during design time. Thresh-

olds for the different colors are set by the company.

Fig. 3 depicts the graphical feedback the manager re-

ceives from MoDe4SLA. A red color indicates the

service is performing worse than agreed upon in the

SLA, while green indicates proper performance of the

service. Yellow indicates the service is not perform-

ing perfectly but still within the boundaries set by the

company, and dark green indicates a service running

even better than the company anticipated. The edges

are colored in the same manner, for example, red indi-

cates the edge was chosen more often than expected.

Fig. 3 indicates that the SLA with the customer is met

regarding costs (i.e., it is colored green); by contrast

the SLA with the customer is regularly violated re-

garding response time, although still within deﬁned

boundaries (i.e., it is colored yellow). Bad perfor-

ICEIS 2009 - International Conference on Enterprise Information Systems

116

mance concerning response time appears to be caused

by content provider 1 (CP1), since its response times

are signiﬁcantly worse than agreed upon. The edges

are colored for their slight deviation from the esti-

mates made at design time. Also the realized impact

factors are different from the estimated ones in Fig. 2.

3 VALIDATION APPROACH

To validate MoDe4SLA we need to determine how

useful experts perceive the use of our feedback mod-

els (cf. Fig. 3). Furthermore we need to determine

whether our approach is effective. The latter applies

when it brings stakeholders closer to their goal, i.e.,

when the approach points to services causing bad per-

formance of the composition. As basis for both ap-

proaches we extend and adopt SENECA (Jaeger and

Rojec-Goldmann, 2005), a simulator for selecting ser-

vices in a composition. In its current state, SENECA

provides the means to generate random composition

structures based on the workﬂow language patterns

by (van der Aalst et al., 2003). In addition, the simu-

lator also generates service candidates with randomly

set QoS values for each candidate. E.g., considered

QoS characteristics are response time, cost, and avail-

ability. In its ﬁrst version, the simulator was designed

to evaluate different selection algorithms for choos-

ing the optimal set of service candidates for the com-

position. SENECA also implements QoS aggrega-

tion method for gaining overall QoS statements of the

composition by aggregatingthe QoS of individual ser-

vices (Jaeger et al., 2004).

3.1 Validation of Usefulness

To evaluate usefulness of our approach we use the

following criterion:

MoDe4SLA is considered as being useful when

experts testing it perceive the feedback given by

MoDe4SLA as more useful for managing and

maintaining the composition than when only using

bilateral monitoring results.

Common management approaches return bilateral

monitoring results to the user. They do not provide

information on the relation between the different ser-

vices but merely return the performance of each indi-

vidual service.

The evaluation setup consists of two parts. On the

one hand we implement the composition simulator to

generate, run, and analyze compositions. On the other

hand we ask a group of experts to test the simula-

tor and afterwards interview them to ﬁnd out whether

they perceive MoDe4SLA as useful.

3.1.1 Simulator

For evaluating the MoDe4SLA approach, the simu-

lator is improved with the ability to perform a dis-

crete event simulation when running the composition

with service candidates. Fig. 5 depicts an overview

of this simulator’s functionality. SENECA randomly

generates a composition structure for a given number

of services. Actual considered structures in the sim-

ulator are sequences, loops, XOR-splits/XOR-joins,

AND-splits and OR-splits (with either a normal join

or a discriminative join). A discriminative join indi-

cates that the structure succeeds when a subset of in-

coming services succeeded. For example, when only

the fastest three out of ﬁve responding services are

considered, then this is a discriminative join.

Each created service in the composition gets as-

signed a randomly generated SLA. The SLA of the

composition is determined by the SLAs of the ser-

vices it depends on. The impact models are derived

using the structure of the composition as well as the

agreed upon service levels. SENECA simulates the

invocation of the candidate services according to the

composition structure. Accordingly, services can fail

with a particular probability, take longer, and, depend-

ing on their use in branching execution paths, differ-

ent cost values will sum up. The simulator will gener-

ate log ﬁles reporting on the runtime results. In addi-

tion, the impact analysis as described in Section 2 is

done and the impact models are generated.

3.1.2 Experts

The experts for our evaluation come from both

academia and industry. We consider people to be

experts if they are familiar with service compositions

and SLAs. We start each session with a training pe-

riod to explain our approach as well as the simulator.

For the actual evaluation we prepare different types of

compositions wrt complexity, i.e., how the structure

is built, number of services, and diversity of SLAs.

For each service monitoring results are gathered.

Two documents for each composition are prepared.

The MoDe4SLA document contains feedback (as

shown in Fig. 3) while the control document contains

the performance data for each service, but does not

provide information on how they are related. Every

expert is ﬁrst offered the control document (cf. 1 in

Fig. 5) and after that the MoDe4SLA document (cf.

2 in Fig. 5). We evaluate the following hypothesis:

MONITORING SERVICE COMPOSITIONS IN MoDe4SLA - Design of Validation

117

SENECA

Generator

Generate

Composition

Generate &

Add SLAs

Generate

Impact Models

Simulator runs

Composition

Gather

Runtime Data

Generate

Feedb. Model

1 2

Figure 5: Evaluating Usefulness.

The MoDe4SLA document has a clear beneﬁt over

the control document for managing the

composition.

For this purpose we will develop a question-

naire addressing several aspects. For example, does

the expert save time by using MoDe4SLA, does

he feel more conﬁdent in making decisions using

MoDe4SLA, and does he ﬁnd the approach helpful.

A typical ﬁve-level Likert item is used to rate the re-

sponse of the experts:

1. Strongly disagree

2. Disagree

3. Neither agree nor disagree

4. Agree

5. Strongly agree

3.2 Validation of Effectiveness

For evaluating effectiveness of our approach we use

the following criterion:

The MoDe4SLA approach is considered to be

effective if managing a service composition using

this approach leads to better results than

managing the composition without it.

We deﬁne better in this case as having better busi-

ness results. In our opinion, this is a twofold process,

where both customer satisfaction as well as proﬁt are

maximized (cf. Table 1). We assume customer sat-

isfaction is high when the number of SLA violations

Table 1: Effectiveness Criteria.

Minimize Maximize

Customer violation costs

Satisfaction average

response time

average costs

Proﬁts service costs customer payments

violation payments violations paid

by providers

is low, costs for the customer are low, and response

time is low. We assume violations being associated

with high penalties have bigger negative impact on

customer satisfaction than low penalties. Therefore,

customer satisfaction is measured through three crite-

ria: payments for violations, average response time,

and average costs for the composition (cf. Table 1).

Maximizing proﬁt is achieved through minimiz-

ing costs and maximizing income (cf. Table 1). In

our case, these costs consist of costs for services and

payments for penalties. Income is generated through

payments by customers and payments by providers

for SLA violations. The total payments for services

to providers and for SLA violations paid to customers

are subtracted from this income.

To evaluate effectiveness, we extend the imple-

mentation described in Section 3.1 so that it becomes

possible to rerun the composition after some manage-

ment choices have been made by the expert. Fig. 6 de-

picts an overview of the effectiveness evaluation ap-

proach. Furthermore, we adapt parts of the service

behavior, we add a renegotiation of SLAs, and ﬁnally

ICEIS 2009 - International Conference on Enterprise Information Systems

118

SENECA

Generator

Generate

Composition

Generate &

Add SLAs

Generate

Impact Models

Simulator runs

Composition

Gather

Runtime Data

Generate

Feedb. Model

Test

Group

Control

Group

Re-Negotiate

SLAs

Re-Negotiate

SLAs

Figure 6: Evaluating Effectiveness.

we evaluate effectiveness of the approach.

3.2.1 Service Behavior

In real life the services in a composition behavediffer-

ently. These differences arise, for example, because

content of services is different, and they are offered

by different providers. The difference in behavior,

together with the structure of the composition, make

it difﬁcult to manage the services. In addition, man-

agers can neither inﬂuence behavior of services pro-

vided by other companies nor can they predict their

behavior other than relying on SLAs. To simulate this

real life complexity, we implement different types of

behavior for services. For example, some services

have low variability in response time, and never vi-

olate the SLA while other services have higher vari-

ability in response time, and violate the SLA more

frequently. Each service gets assigned a behavior type

at design time. Which type of behavior is assigned to

the service, is unknown to the expert.

To realize this, we implement agents to steer the

different services in a composition. Each agent gets

assigned a behavior type, for example, the reliable be-

havior type. Behavior of the service depends on both

behavior type and agreed upon SLA. For example, if

the behaviortype is reliable and agreed upon response

time is lower than 10 ms, a random distribution of re-

sponse time will be generated for the next 100.000

invocations. This distribution is created with the pa-

rameters reliable and lower than 10 ms. At runtime,

with each invocation of the service, a response time

is randomly chosen from this distribution. As a re-

sult each service shows different behavior, ﬁtting its

SLA. Some services might violate their SLAs more

severely, or more frequently than other services be-

cause of their assigned behavior type.

3.2.2 Renegotiation

After running the composition the expert is asked to

renegotiate SLAs for a subset of the services. In prac-

tice, determining which SLAs to renegotiate is par-

ticulary difﬁcult for complex compositions, i.e., com-

positions with many services and constructs. Since

MoDe4SLA graphically pinpoints to badly perform-

ing services and, moreover,indicates the impact these

services have on the composition, it becomes possible

to prioritize SLAs, decreasing management efforts.

As described, services in the simulator expose

unique behavior: some perform with only minor de-

viation in, for example, response time while others

ﬂuctuate more. In a perfect agreement, the SLA be-

tween provider and customer reﬂects this behaviorex-

actly. In practice, however, it is necessary to monitor

service behavior and renegotiate the SLAs for badly

performing services. In these negotiations provider

and customer have conﬂicting interest since both aim

at maximizing monetary beneﬁts. To simulate this in

the evaluation, the experts are asked to select a subset

of services for renegotiation after the ﬁrst run. Each

service agent offers three new possible SLAs to the

expert. E.g., the original SLA has fast response time

but low violation penalties. A newly offered SLA by

the agent might have slower response time, but high

penalties when being violated. These new SLAs are

MONITORING SERVICE COMPOSITIONS IN MoDe4SLA - Design of Validation

119

generated taking the behavior type of the agent as well

as the original SLA into account. Some of the new

SLAs will ﬁt better to the service from the customer

perspective and some will ﬁt the service better from a

provider perspective. The challenge for the expert is

to make choices beneﬁcial for his company.

3.2.3 Evaluation

We use one test group and one control group for

each tested service composition. The test group uses

MoDe4SLA to choose services they want to rene-

gotiate, while the control group uses log ﬁles only.

MoDe4SLA is designed to assist the user in making

choices on which services to renegotiate. Further-

more, MoDe4SLA pinpoints to the exact problems

of the service so that choices between newly offered

SLAs are easier to make. If the expert successfully

identiﬁes services with high impact on the compo-

sition and which are performing badly according to

their SLAs, renegotiation will have to result in bet-

ter runtime results of the test group than the control

group (see criteria summarized in Table 1).

4 RELATED WORK

(Sahai et al., 2002) aim at automated SLA moni-

toring by specifying SLAs and not only considering

provider side guarantees but focus also on distributed

monitoring, taking the client side into account. (Bar-

bon et al., 2006) enable run-time monitoring while

separating the business logic from monitoring func-

tionality. For each instance of a process a monitor

is created. Unique in this approach is the ability to

also monitor classes of instances, enabling abstrac-

tion from an instance level. The smart monitoring ap-

proach of (Baresi et al., 2004) implements the moni-

tor itself as service. There are three types of monitors

available for different aspects of the system. Their ap-

proach is developed to monitor speciﬁcally contracts

with constraints. (Baresi and Guinea, 2005) presents

an approach to dynamically monitor BPEL processes

by adding monitoring rules to the different processes.

These rules are executed during runtime. Our ap-

proach does not require modiﬁcations to the process

descriptions what suits better to some application ar-

eas. An interesting approach in this direction is work

by (Mahbub and Spanoudakis, 2005) which, as en ex-

ception, do consider the whole state of the system in

their monitoring approach. They aim at monitoring

derivations of behavior of the system. The require-

ments for monitoring are speciﬁed in event calculus

and evaluated with run-time data. Although many

of the above mentioned approaches do consider third

parties and allow abstraction of results for composite

services, none of them addresses how to create this

abstraction in detail. E.g., matching messages from

different processes as in our SubscribedNews exam-

ple where databases are used, are not considered.

(Menasce, 2004) presents an analysis for the re-

sponse time of composed services with the goal to

identify the impact of slowed down services. The

impact on the composition is computed by using a

Markov chain model. The result is a measure for the

overall slow down depending on the statistical like-

lihood of a service not delivering the expected re-

sponse time. As opposed to our approach, Menasce’s

approach performs at design-time rather than provid-

ing an analysis based on monitoring runtime data. In

addition, our work provides a framework a) to cover

structures beyond an fork-join arrangement, and b)

that supports different measures subject of an SLA

in addition to response time. A different approach

with the same goal is the virtual resource manager

proposed by (Burchard et al., 2004). This resource

management targets a grid environment where a cal-

culation task is distributed among different grid nodes

for individual computation jobs. The organization of

the grid is hierarchically separated into different ad-

ministrative domains of grid nodes. If one grid node

fails to deliver the assured service level, a domain

controller ﬁrst reschedules the job onto a different

node within the same domain. If this fails, the domain

controller attempts to query other domain controllers

for passing the computation job. Although the goal

is the same and the approach covers runtime, it fol-

lows a hierarchical autonomic recovery mechanism.

MoDe4SLA focusses on identifying causes for cor-

rection on the level of business operations rather than

on autonomously performed job scheduling.

Another research community analyzes root causes

in services. In corresponding approaches depen-

dency models are used to identify causes of violations

within a company. Here, composite services are not

considered but merely services running in the com-

pany. E.g., (Agarwal et al., 2004) determine the root

cause by using dependency graphs. Especially ﬁnd-

ing the cause of a problem when a service has an

SLA with different metrics is here a challenging topic.

Also (Caswell and Ramanathan, 2000) use depen-

dency models for managing internet services, again,

with focuss on ﬁnding internal causes for problems.

MoDe4SLA identiﬁes causes of violations in other

services rather than internally. Furthermore, our de-

pendencies between different services are on the same

level of abstraction while in root cause analysis one

service is evaluated on different levels of abstraction.

ICEIS 2009 - International Conference on Enterprise Information Systems

120

5 SUMMARY AND OUTLOOK

In this paper we describe our plan to evaluate

MoDe4SLA which is intended to support the com-

pany in managing its composite services for detec-

tion and coverage of SLA violations. We propose

to do a twofold evaluation where ﬁrst usefulness of

MoDe4SLA as perceived by experts is tested after

which effectiveness is evaluated by actually managing

a service composition using MoDe4SLA. Both steps

are supported by SENECA, a simulator which is ex-

tended for this purpose.

Currently, we are in the process of evaluating the

usefulness of our approach. In near future we plan to

extend our current implementation to also evaluate ef-

fectiveness. Furthermore, we extend our dependency

based approach with handling complex SLAs where

also dependencies between and within different SLOs

occur, e.g., SLAs where two different SLOs concern

response time, or where one SLO concerns not only

response time but also cost.

REFERENCES

Agarwal, M. K., Appleby, K., Gupta, M., Kar, G., Neogi,

A., and Sailer, A. (2004). Problem determination us-

ing dependency graphs and run-time behavior models.

In DSOM, volume 3278, pages 171–182. Springer.

Barbon, F., Traverso, P., Pistore, M., and Trainotti, M.

(2006). Run-time monitoring of instances and classes

of web service compositions. In ICWS ’06, pages 63–

71, Washington, DC, USA.

Baresi, L., Ghezzi, C., and Guinea, S. (2004). Smart moni-

tors for composed services. In ICSOC ’04, pages 193–

202, New York, NY, USA.

Baresi, L. and Guinea, S. (2005). Towards dynamic moni-

toring of WS-BPEL processes. In ICSOC 05’, pages

269–282.

Bodenstaff, L., Wombacher, A., Reichert, M., and Jaeger,

M. C. (2008). Monitoring dependencies for SLAs:

The MoDe4SLA approach. In SCC (1), pages 21–29.

Burchard, L.-O., Hovestadt, M., Kao, O., Keller, A., and

Linnert, B. (2004). The virtual resource manager:

an architecture for SLA-aware resource management.

Cluster Computing and the Grid, pages 126–133.

Cardoso, J., Sheth, A., Miller, J., Arnold, J., and Kochut,

K. (2004). Quality of service for workﬂows and web

service processes. Web Semantics: Science, Services

and Agents on the World Wide Web, 1(3):281–308.

Caswell, D. and Ramanathan, S. (2000). Using ser-

vice models for management of internet services.

IEEE Journal on Selected Areas in Communications,

18(5):686–701.

Jaeger, M. C. and Rojec-Goldmann, G. (2005). SENECA -

simulation of algorithms for the selection of web ser-

vices for compositions. In TES, pages 84–97.

Jaeger, M. C., Rojec-Goldmann, G., and M¨uhl, G. (2004).

QoS aggregation for service composition using work-

ﬂow patterns. In (EDOC’04), pages 149–159.

Mahbub, K. and Spanoudakis, G. (2005). Run-time moni-

toring of requirements for systems composed of web-

services: Initial implementation and evaluation expe-

rience. In ICWS ’05, pages 257–265.

Menasce, D. A. (2004). Response-Time Analysis of Com-

posite Web Services. In IEEE Internet Computing,

pages 90–92. IEEE Press.

Sahai, A., Machiraju, V., Sayal, M., van Moorsel, A. P. A.,

and Casati, F. (2002). Automated SLA monitoring for

web services. In DSOM ’02, pages 28–41, London,

UK. Springer-Verlag.

Tosic, V., Pagurek, B., Patel, K., Esfandiari, B., and Ma, W.

(2005). Management applications of the Web Service

Offerings Language (WSOL). Information Systems,

30(7):564–586.

van der Aalst, W., ter Hofstede, A., Kiepuszewski, B., and

Barros, A. (2003). Workﬂow Patterns. Distributed

and Parallel Databases, 14(1):5–51.

Zeng, L., Benatallah, B., Ngu, A. H., Dumas, M.,

Kalagnanam, J., and Chang, H. (2004). QoS-aware

middleware for web services composition. IEEE

Trans. Softw. Eng., 30(5):311–327.

MONITORING SERVICE COMPOSITIONS IN MoDe4SLA - Design of Validation

121