Balancing Functionality, Risk, and Cost in Smart Service Networks

F. Rizal Batubara

Services, Cyber Security and Safety Research Group, University of Twente, Enschede, The Netherlands

1 CONTEXT

Recent developments in the area of software service

systems brought a new level of scale, complexity

and pervasiveness to cope with. This has, in turn,

changed the business model of companies engaged

in the software industry and the way they present

their services. Healthcare systems, as an example,

transform services from a traditional professional-

centric care model to model that is characterized as

‘pervasive healthcare’. Pervasive healthcare applies

extramural networked healthcare systems with

sensors and devices at the patients’ location, making

the patient an active partner in the care process.

These transformations of technologies and business

models have brought a new breed of systems called

Smart Service Networks (SSN). An SSN involves

any number of devices, sensors and IT systems, and

has diverse stakeholders, which together form a

network in which resources are integrated and

applied through interaction (Stroulia, 2010).

An SSN is a context-aware system, it can adapt

and provide specific services to the user according to

the context information received from data collected

with sensors. Therefore, SSN must be able to clearly

understand the significance of the context

information conveyed from its environment. The

quality of the context model determines the extent to

which the SSN can offer services that fit the actual

context, which in turn determines the usefulness of

the offered services to the user. The quality of the

context model depends on the quality of the

collected context data (accuracy, timeliness, etc.).

Low(er) data quality means low(er) model quality,

which can lead to (more) off-topic service offerings.

Off-topic service offerings can be useless to the user,

and may even have negative value to the user.

SSN development projects operate in a multi-

stakeholders context and potentially experience

conflicts among functionality, risk and cost to meet

the stakeholders’ requirements. This research

focuses on trading-off functionality, risk and cost

within SSN development. By defining the possible

trade-off scenarios in SSN delivery projects, this

research will provide a rational analysis to justify

SSN design decisions that lead to achieving software

systems with adequate functionality, minimum risk

and reasonable cost.

2 RESEARCH PROBLEM

Developing a SSN is a valuable business investment

decision that is expected to bring returns from its

investments (Van Den Heuvel and Papazoglou,

2010). Surely the stakeholders expect the system

with as complete as possible functionalities and as

high as possible quality, to earn profit for them.

While meeting the required functionality, the

developers must also meet safety standards and

comply with cost estimates based on risk analysis.

Strategic decisions during the design of a SSN need

a joint control of the estimated functionality, risk

and cost.

The relationship between functionality, risk, and

cost for SSN development (or any IT system for that

matter) is described in Figure 1. Therein, the boxes

represent variables of interest – in our case

functionality, risk, and cost – and the directed

connections represent cause-effect relationships. The

‘+’ sign with a connection means that the connected

variables change value in the same direction; the ‘-‘

sign means that the connected variables change

value in opposite directions.

Figure 1: Relationships between functionality, risk, and

cost in smart service network development.

We make the note that in Figure 1, the concept of

functionality and risk reflect the understanding that

in any SSN, there are ’on-topic’ SSN services

represent value-added functionality and ‘off-topic’

Batubara F..

Balancing Functionality, Risk, and Cost in Smart Service Networks.

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

SSN services represent risks. Value-added

functionality may be enhanced by considering more

elaborate context models. However, embedding

those models into the SSN would contributes to

increased cost. Meanwhile, risks can be mitigated by

enhancing the quality of context models and/or only

offering services if a threshold quality is satisfied,

this contributes to costs as well.

Based on these phenomena in the context of

SSN, the problem statement of this research is:

To design a method to balance trade-off among

functionality, risk, and cost in order to support

decisions on adequate functionality, minimum risk

and reasonable cost within smart service network

development.

Following the methodology of (Wieringa, 2014),

we formulated the following research questions

(RQ) which address the problem stated:

1. What is empirical evidence about relationship

among functionality, risk, and cost in a smart

service network, and what mechanisms explain

this relationship? What could be a working

definition of an ‘optimal solution’ from client

and vendor perspective regarding SSN systems??

RQ 1.1: What definitions of functionality, risk

and cost are useful in the context of

SSN?

RQ 1.2: What empirical evidence exists about

the possible relationships among these

three concepts?

RQ 1.3: Are there any assumptions about the

relationships indicated in published

literature sources?

RQ 1.4: How can we explain these relationships

in terms of mechanisms describing a

SSN operating in a context?

RQ 1.5: What is an ’optimal balance’ among

these three concepts, according to

published empirical evidence?

2. How to achieve an optimal solution in order to

maximize the functionality, minimize the risk,

and minimize the cost within the development of

SSN?

RQ 2.1: What methods are available to estimate

functionality, risk, and cost within SSN

development? What are the

requirements for such a method? Which

method should we use?

RQ 2.2: What approaches, theories, and models

allow us to balance the trade-off among

these three concepts?

RQ 2.3: What is the model to calculate and what

are the criteria to decide on an optimal

balance of these three concepts?

RQ 2.4: How to enhance the SSN design process

to incorporate the proposed method for

balancing between these three concepts?

RQ 2.5: How can a SSN architecture support this

enhanced design process, e.g. by

facilitating traceability between

balancing decisions and architecture

components?

3. How to validate the proposed solution approach

to balance among functionality, risk, and cost in

SSN system?

RQ 3.1: Can we apply (a prototype of) our

method in a setting that allows

validation of its properties?

RQ 3.2: What are the effects of applying our

method?

RQ 3.3: Do these effects satisfy the requirements

that were put forward on the method –

optimizing the balance between

functionality, risk, and cost?

RQ 3.4:

What can we say about the

dependencies of the observed effects on

the chosen setting, and about the

generalizability of the observed effects

over different settings?

3 OBJECTIVE

The goal of this PhD research is to design a method

for balancing functionality, risk and cost in SSN

development projects. This method will be applied

to support stakeholders in resolving the trade-off

problems that are common for SSN development,

thereby helping produce SSN software with

adequate functionality, minimum risk and

reasonable cost.

4 STATE OF THE ART IN SMART

SERVICE NETWORK

Integration of legacy and new software systems

through service oriented architecture (SOA) and

cross-platform standards opens up opportunities for

new services that will endorse smart services in

various domains. The integration yields many

benefits, with new software systems (and associated

technologies) enabling a transformation from

traditional services to smart services (OECD, 2013).

To increase value for clients, several SSN can

integrate existing smart services in order to deliver

new services that accommodate the co-production of

ICSOFT2015-DoctoralConsortium

new knowledge and services through organic peer-

to-peer interactions (Wang et al., 2012). Customers

and partners who participate in this network can be

mutually providing benefits. Hence, SSN utilizes

network services with distributed computing, such

as sensor network technologies. They interweave

smart physical devices, actively monitoring,

analyzing and notifying consumers about everyday

service needs (Van Den Heuvel and Papazoglou,

2010).

4.1 Characteristics

In order to establish and develop a new discipline of

services such as made possible by a SSN, revealing

the SSN’s essential characteristics is of paramount

importance. Understanding these characteristics will

be instrumental to formulate the definition of SSN

comprehensively. Plus, from a practical point of

view, it will help us establish appropriate models of

SSN.

In early research, many researchers have

revealed some characteristics of SSN and defined

various notions of SSN (OECD, 2013);

(Allmendinger and Lombreglia, 2005); (Wang et al.,

2012); (Van Den Heuvel and Papazoglou, 2010);

(Dos Santos Pacheco et al., 2013); (Stroulia, 2010)

(Ng et al., 2010). Yet there is still no widely

accepted definition of SSN.

Organization for Economic Co-operation and

Development (OECD) defined SSN as the result of a

combination of three distinct phenomena, namely

machine to machine communication to transmit data,

cloud computing to process and display the data, and

big data analysis to correlate and interpret the data

(OECD, 2013). The expansion of machine to

machine communication will enable interconnected

devices to form an SSN and produce large volumes

of data. Large scale processing will be delivered by

cloud computing services and analysis of these data

will be undertaken around a process frequently

called “Big Data”.

(Wang et al., 2012) and (Van Den Heuvel and

Papazoglou, 2010) considered SSN as “systems of

service systems that are open, complex and fluid,

accommodating the co-production of new knowledge

and services through organic peer-to-peer

interactions”. “Open” in this case means that the

system involves a wide range of systems to improve

service innovation in an increasingly complex and

dynamic environment. “Complex” relates to a joint

effort of interdisciplinary collaboration, cooperation

and coordination among the network participants.

Further, “liquid” means the system should have

creative approaches where services and interactions

are implemented with awareness and dynamic

adaptation to the users’ computational environments,

changing policies and unknown requirements.

Another characteristic of SSN is that it relies on an

IT infrastructure that enables network partners to

seamlessly integrate their software services with

back-end systems and wireless sensor networks.

(Stroulia, 2010) associated SSN with the system

attributes: instrumented, interconnected, and

intelligent. The term “instrumented” refers to the

embedding of “sensing devices” in the environment

where the service-delivery process occurs.

“Interconnected” is contributed by large-scale

instrumentation of the world around us including

data, system and people interconnectedness. Finally,

the term “intelligent” refers to the numerous

analyses that are possible on the multitude of

information available and based on which the system

can inform and improve our decision making.

In addition, (Villegas and Müller, 2010)

explained that “a system should be able to reason

about its current state and provide functionality

based on its context and the current user’s matters

of concern in order to support smart interaction and

smart services”, so that the system has to be

reflective and context-aware.

Drawing on the literature sources in this section,

we conclude that:

▪ SSNs have the ability to construct and maintain

a model of their dynamic context, and to

proactively offer services to the user when

these services satisfy corresponding criteria

regarding the context model state;

▪ SSNs may have the ability to learn about new

criteria (context patterns) for offering services;

▪ SSNs use sensing devices, machine to machine

communication, and reasoning systems to

support the above-mentioned context-

awareness and learning;

SSNs consist of many different subsystems from

different manufacturers and makes, and therefore

should be based as much as possible on open

standards.

4.2 Definition

In preparing for this research, we did a preliminary

literature survey, by searching relevant articles on

SSN in electronic libraries (ACM Digital Library,

IEEE Xplore, ISI Web of Science, Scopus and

SpringerLink). Our survey found only a few

definitions of SSN. Also, we found that SSN

BalancingFunctionality,Risk,andCostinSmartServiceNetworks

terminology covers many synonyms as terms, and it

is hard to nail down unique words to define the same

concept. However, all these concepts are not in

contradiction with each other, as they share some

commonalities in their meanings and are partially

overlapping. We think we could leverage these

concepts and map them against the SSN

characteristics, if we want to come up with a more

specific definition of SSN. Table 1 provides

examples of definitions we found in literature.

Table 1: SSN definitions from literature.

Source

Definition

OECD, 2013

An application or service that is

able to learn from previous

situations and to communicate the

results of these situations to other

devices and users.

Wang et al., 2012

Open systems accommodating the

co-production of new knowledge

and services through organic peer-

to-peer interactions.

Van Den Heuvel and

Papazoglou, 2010

Leveraging service networks with

novel distributed computing, such

as sensor network technologies.

Allmendinger and

Lombreglia, 2005

A brand new configuration of

services that are fundamentally

pre-emptive rather than reactive or

even proactive. Aggressive actions

taken by people based on

intelligence to prevent undesirable

results. For customers, smart

services prevent unpleasant

surprise in their lives.

As we could see, the definitions in Table 1 do

not describe SSN completely because they do not

yet include the characteristics of SSN as identified

previously. By analyzing the existing definitions and

identified SSN characteristics, SSN can be possibly

described as suggested in Figure 2.

Figure 2: Smart service network (straight-line arrows

mean compositions and broken-line arrows mean added

capabilities).

Furthermore, for the purpose of this research, we

propose the following comprehensive definition of

the SSN:

“A Smart Service Network is a dynamic system

which combines smart services and it offers services

in accordance its context and user needs, using

sensing devices, machine to machine

communication, and reasoning systems”

For example, let’s consider a pervasive

healthcare system which is able remotely to monitor

a patient’s vital parameters, such as

electrocardiography (ECG), blood pressure and

oxygen saturation level. The patient is equipped with

ambulatory sensors to acquire patient’s health data,

which is subsequently transmitted to a local PC

station at the patient’s home. The local PC runs

special application to analyse the acquired data and

produces alerts, according to the parameter threshold

table (the doctor is able to remotely set the threshold

values of the monitored parameters on the patient’s

PC). In case of any abnormality the findings and the

collected vital data are immediately transmitted to

the clinic for further analysis. At the clinic, the

monitoring module is responsible for monitoring the

patient’s progress. It encapsulates the rules for

manipulating the database and the parametric model

of the typical patient, for each disease. Then, using

rules related to the monitoring of the patient’s

progress, the system assigns specific values to the

parameters of the model, based on the readings of

the patient’s sensors.

In this example, two smart services (patient’s

monitoring and clinical analysis) are combined

together to produce a pervasive healthcare service

that can provide better recommendations for

patient’s health. This system may involve some

providers such as hospitals, laboratories,

pharmacies, and insurance companies. Patients use

some sensors and are connected to the system

through some communication lines. This healthcare

system has the ability to learn from previous

patients’ health records and context awareness based

on patient’s vital parameters collected, so it can be

classified as SSN.

In this case, if we add new functionality like auto

recommendation of drugs to patient, it will cost

more on system development and also stimulate a

new risk if the recommendation that was produced is

wrong. Meanwhile, if we want the system to be safe

to use, maybe we have to reduce some

functionalities that potentially may cause error in

recommendations. These are examples of the trade-

off that may occur in the SSN.

ICSOFT2015-DoctoralConsortium

4.3 Related Works

In the disciplines of Requirements Engineering and

Software Engineering, contradicting or conflicting

quality requirements are treated from three

perspectives: requirements negotiation, requirements

conflict resolution and requirements prioritization

(Herrmann and Daneva, 2008). While all

perspectives do address trade-offs between quality

attributes, the ways through which balance is

achieved differs. In turn, there are numerous

approaches to reconciling conﬂicts between aspects

of software product quality. Some of the most

common include expert judgment, Quality Function

Deployment (QFD) and Theory W (Henningsson

and Wohlin, 2002). The literature reviewed depicts a

number of methods and techniques for use in dealing

with software quality trade-offs.

However, there are different opinions as to the

source of conflict dependencies. Some authors have

stated that conflict is inherent to a pair of quality

requirements, while others emphasize that conflict

interactions depend on the software architecture and

coding (García-Mireles et al., 2013). Both opinions

may be true for different requirements of the system.

For example, delay and reliability are conflicting,

but software architecture and technology choices

help to get the required balance of performance. The

most common methods reported to support software

quality trade-off are the Analytical Hierarchy

Process (AHP), model building, the Architecture

Trade-off Analysis Method (ATAM), algorithm-

based and metric-based methods, expert opinion,

Quality Function Deployment (QFD) and prototypes

(García-Mireles et al., 2013) (Barney et al., 2012).

Additional techniques exist such as goal models and

the automated construction of architecture

alternatives. However, little empirical support is

provided to prove their effectiveness and usefulness

when software quality trade-offs are involved.

Furthermore, in the discipline of Software

Architecture, a broad range of software architecture

trade-off methods have also been proposed and

evaluated in order to understand the benefits and

shortcomings of each method. (Falessi et al., 2011)

compared decision-making techniques at the

software design stage, taking into account the

difficulties involved in using it. They found that

there is no universally best decision-making

technique for the resolution of trade-offs in

architecture design, it depends on which difficulties,

issues, or troubles the architects want to avoid.

(Barney et al., 2012) did a systematic map of the

software quality trade-off literature to understand the

state of the research addressing this area by

systematic review. They found that there is an

immature approach of software quality trade-off

have emerged as candidates to dominate the research

space. Only 28% of 168 relevant publications that

have been reviewed provide empirical evidence. At

61% of the research provide non-empirically

assessed solution proposals. Accordingly, there is no

clear approach used for software quality trade-offs.

Nevertheless, there is no approach that I have

found to do trade-off analysis among functionality,

risk, and cost for SSN. SSN need its own trade-off

analysis since it has different characteristics from

other systems such as it involves many service

providers and clients, can be combined with

individual units that are smart, and use a lot of

sensors, actuators and communication.

5 METHODOLOGY

To meet the research objective, we will use the

principles and guidelines of design science

methodology by (Wieringa, 2014) to answer our

research questions. It includes four main parts:

implementation evaluation/problem investigation,

treatment design, design validation and treatment

implementation, as shown in Figure 3.

Figure 3: Research methodology used in this research

(Wieringa, 2014).

In the context of the work will be conducted in this

research, we will perform the following actions for

each part:

1. Problem investigation: we have to identify the

stakeholders, goals, phenomena and

contributions of this research in order to gain a

deeper understanding of the problem to be

solved. Furthermore, we need to define concepts

of smart service network, functionality, risk, cost

and relationship among them to get clear

objective about the problem definition. These

investigations will be carried out by surveying

existing literature through systematic literature

BalancingFunctionality,Risk,andCostinSmartServiceNetworks

review (Kitchenham and Charters, 2007).

Observational case studies also needed in this

steps to capture mechanisms that produce real-

world phenomena. Based on the problem and

domain investigation we formulate the criteria

and requirements for our solution, which

contribute to the stakeholders’ goals. Next, we

have to determine the best method that suitable

with SSN for estimating functionality, risk and

cost to be used in calculation to balance the

trade-off. This task will be conducted by doing a

literature review and accompanied by

observational case studies.

2. Treatment design: based on the result of the

problem investigation, we will formulate the

objective function to calculate the relationship

among functionality, risk and cost. This

objective function will generate a model to

simulate the trade-off optimization. Then based

on the simulations and evaluation of the model,

we will propose a method to optimize the trade-

off that allows us to determine adequate

functionality, minimum risk, and reasonable cost

provided. In this part, we also need to find out

how to use the proposed method to enhance SSN

design process.

3. Design validation: to validate the method, we

will ask expert opinion and use SSN project in

companies to demonstrate the usefulness and the

utility of the proposed method. Expert opinion is

the simplest way to validate an artefact by asking

them about our proposed method whether it can

resolve the trade-off problem. If the predicted

results do not satisfy our goal, then we have to

redesign the method. Therefore, we will use this

method in real-life settings to obtain a real

picture of the application of this method in the

real SSN development project. The lessons

learned in this validation will serve to formulate

improvements to the method.

4. Treatment implementation: this task is the start

of an implementation process to hand over the

method to practitioners. This is not part of our

research project.

6 EXPECTED OUTCOME

The main expected outcome of this research is

knowledge about how to balance the trade-off

among functionality, risk and cost required within

SSN development. A secondary expected outcome

lies in exploring the relationship of functionality,

risk and cost in the domain of SSN. Furthermore, by

analyzing existing available estimation method, their

suitability with SSN and maybe with slight

modifications or adjustments, we could propose the

best method for estimating such things. Finally, by

defining objective function for optimal trade-off and

simulating with a model then we can produce a

method for balancing the trade-off. This method will

allows SSN providers to solve the trade-off

problems and provides services with adequate

functionality, minimum risk, and reasonable cost.

7 STAGE OF THE RESEARCH

This research will be carried out through four stages

as described in Figure. Furthermore, steps to be

performed to answer the research questions can be

explained in more detail as follows:

1. Problem and domain investigation

Problem and domain investigation will answers

RQ1.1 to RQ2.2 about definitions of

functionality, risk and cost, relationships and

assumptions of the trade-off, the best method

that suitable with SSN for estimating

functionality, risk and cost, and optimal balance

concept. This stage will be done by literature

review and observational case studies.

2. Design a new method

To answer RQ2.3 and RQ2.5, we need to

formulate criteria to decide on an optimal

balance. Based on these criteria a model to

simulate the trade-off optimization will be

generated. Then a method to optimize the trade-

off will be proposed and we will try to define an

SSN design process enhancement within the

method proposed.

3. Validation

Finally, to answer RQ3.1 and RQ3.2 we will ask

expert software engineers to express their

opinion about the method in focus groups

discussion. The next validation task is to use this

method in real projects such as Smart Reasoning

Systems For Well-Being at Work and at Home

(SWELL) project. Lessons learned from these

projects will be used to improve and finalize the

method further.

We started this research with a literature review to

get familiar with the state-of-the-art and to find

characteristics and clear definition of SSN. Based on

this review we have found some characteristics of

SSN and tried to define SSN comprehensively.

Moreover, from reviewed literature, we found that

there is no approach exist to do trade-off analysis

ICSOFT2015-DoctoralConsortium

Figure 4: Methodological structure (The boxes mean tasks, white arrows mean task decompositions, and numbers indicate

the sequence of tasks will be performed. The black arrows connecting boxes at the left side of the figure indicate sequence

of tasks in the design cycle and will be executed from top to bottom).

among functionality, risk, and cost for SSN. SSN

need its own trade-off analysis since it has different

characteristics from other systems such as it has

ability to construct and maintain a model from

dynamic context, has ability to learn, and uses

sensing devices, machine to machine

communication and reasoning system. These

characteristics cause the SSN becomes very risky,

could even endanger the user if the

recommendations produced was wrong.

For next step, we will assess the form of

relationship among functionality, risk, and cost and

to define an ‘optimal solution’ for the trade-off in a

SSN development. To assess this relationship, we

will perform a literature review and observational

case studies to validate the results.

8 CONCLUSION

Due to the complexity, heterogeneity, and

dynamism, the development of SSN may potentially

BalancingFunctionality,Risk,andCostinSmartServiceNetworks

be fraught with conflicts among functionality, risk

and cost. Reasoning about those conflicts and

resolving them effectively is an important part of

achieving a well-managed SSN development

process. This research focused on designing a

method to balance functionality, risk and cost in

SSN development projects. The method is expected

to help the stakeholders to choose and decide on

balanced functionality, risks and costs according to

their needs. We provided our first working definition

of SSN based on a literature survey. We motivated

the need for more research in the area and defined

our research questions and research process by using

design science principles. Our research methodology

and planned research stages will translate into more

specific research activities as our research is

progressing. We started only recently and in the next

months we expect .many concepts to become clearer

and new challenges to crystallize while the research

is advancing.

REFERENCES

Allmendinger, G. & Lombreglia, R. 2005. Four strategies

for the age of smart services. Harvard business

review, 83, 131-+.

Barney, S., Petersen, K., Svahnberg, M., Aurum, A. &

Barney, H. 2012. Software quality trade- offs: A

Systematic Map. Information and software

technology, 54, 651-662.

Dos Santos Pacheco, R. C., Salm Junior, J. F., Koller, k. &

Schneider, V. 2013. Knowledge and ontology

engineering for smart services.

Falessi, D., Cantone, G., Kazman, r. & Kruchten, P. 2011.

Decision- making techniques for software architecture

design: a comparative survey. Acm computing surveys

(csur), 43, 1-28.

García-mireles, G., Moraga, M., García, F. & Piattini, M.

2013. A Framework to support software quality trade-

offs from a process-based perspective. In: Mccaffery,

f., O’Connor, r. & Messnarz, r. (eds.) Systems,

software and services process improvement. Springer

berlin heidelberg.

Henningsson, K. & Wohlin, C. Understanding the

relations between software quality attributes – a

survey approach. 12

international conference on

software quality, 2002 ottawa, canada. Asq software

division.

Herrmann, A. & Daneva, M. 2008. Requirements

prioritization based on benefit and cost prediction: an

agenda for future research.

Kitchenham, B. A. & Charters, S. 2007. Guidelines for

performing systematic literature reviews in software

engineering. Technical report ebse-2007-01. Version

2.3 ed. Keele, uk: ebse.

NG, J. W., Chignell, M., Cordy, J. R. & Yesha, Y. 2010.

Smart services. Lecture notes in computer science

(including subseries lecture notes in artificial

intelligence and lecture notes in bioinformatics), 6400,

173-177.

OECD 2013. Building blocks for smart networks. Oecd

digital economy papers. OECD.

Stroulia, E. 2010. Smart services across the real and

virtual worlds. Lecture notes in computer science

(including subseries lecture notes in artificial

intelligence and lecture notes in bioinformatics), 6400,

178-196.

Van den heuvel, w. J. & papazoglou, m. P. 2010. Toward

business transaction management in smart service

networks. IEEE internet computing, 14, 71-75.

Villegas, N. M. & Müller, H. A. 2010. Managing dynamic

context to optimize smart interactions and services.

Lecture notes in computer science (including subseries

lecture notes in artificial intelligence and lecture notes

in bioinformatics), 6400, 289-318.

Wang, Y., Taher, Y. & Van den Heuvel, W. J. 2012.

Towards smart service networks: An interdisciplinary

service assessment metrics.

Wieringa, R. J. 2014. Design science methodology for

information systems and software engineering.

Springer.

ICSOFT2015-DoctoralConsortium