=Paper=
{{Paper
|id=Vol-1216/paper6
|storemode=property
|title=Quality in Use and Software Greenability
|pdfUrl=https://ceur-ws.org/Vol-1216/paper6.pdf
|volume=Vol-1216
|dblpUrl=https://dblp.org/rec/conf/re/CaleroMBD14
}}
==Quality in Use and Software Greenability==
https://ceur-ws.org/Vol-1216/paper6.pdf
Quality in Use and Software Greenability
Coral Calero, Mª Ángeles Moraga Manuel F. Bertoa Leticia Duboc
Instituto de Tecnologías y Sistemas de Departamento de Lenguajes y Ciencias Universidade do Estado do Rio de
la Información de la Computación Janeiro
University of Castilla-La Mancha University of Malaga Rio de Janeiro, Brasil.
Ciudad Real, Spain Malaga, Spain leticia@ime.uerj.br
{Coral.Calero; bertoa@lcc.uma.es
MariaAngeles.Moraga}@uclm.es
Abstract— Software sustainability has recently begun to gain However, none of these models considers sustainability or
importance. However, although proposals concerning what it is the ecological aspects of software products. This is, from our
and how to achieve it are starting to appear, until now, there point of view, an important weakness of the standards, since
have been very few proposals on how to model it. Sustainable sustainability is gaining more and more importance in society
development consists of three dimensions: Social sustainability,
in general and in industry in particular. We believe this
Economic sustainability and Environmental sustainability; the
latter being more closely related to technical aspects. There are characteristic should also be considered in the software quality
also three environmental impact levels for ICTs: direct context. Our proposal is to complete the quality models of the
environmental effects of production and use of ICTs, indirect standard in such a manner that we will be able to take
environmental impacts related to the effects of ICTs and indirect sustainability into consideration when developing or
effects on the environment. In this paper, we focus on evaluating a software product. This will allow to create the
environmental sustainability and the first environmental impact necessary foundation to incorporate into and assess
level, and more concretely, on the direct environmental effect of greenability of a software product. In this paper, we present
software use. We specifically propose a greenability in use our progress with respect to this goal.
characteristic to be considered as part of the quality in use model
The reminder of this paper is organized as follows: Section
proposed by the ISO 25010 standard. This model can be used
using measures, indicators, or even Bayesian Networks in order two presents important aspects of sustainability in general.
to link it with product quality. We therefore present an example Section three discusses the importance of sustainability in the
of a Bayesian Network that links product quality to greenability software context and presents the focus of our research.
in use. Our eventual goal is to provide developers with indicators Section four describes the relevant aspects of the ISO/IEC
and guidelines on how to develop an environmentally friendly 25010 standard and its quality models. The product
software product. greenability characteristic, proposed by the authors in a
previous work, is also shown in the fourth section. In Section
Index Terms—Software quality, green software, greenability five, we propose the greenability in use, a new characteristic
I. INTRODUCTION to be added to the quality in use model of the standard. Some
examples of the levels of impact of this new characteristic are
Quality is currently among organizations’ main goals shown in Section six. Section seven shows how to combine
because nowadays the industry made efforts to obtain the the product quality and the quality in use models by means of
ISO9000 or CMMI (even been mandatory in the USA if a Bayesian networks. Finally, Section eight presents our
company want to collaborate with the government). The conclusions and future work.
SWEBOK (Software Engineering Body of Knowledge) [16],
the main reference guide of software engineers includes a II. SUSTAINABILITY
specific chapter about software quality and how to apply it to
One of humanity’s current challenges is to conserve the
the software engineering discipline. A large number of environment and attain a sustainable economic and social
organizations provide similar products, thus permitting development.
consumers to choose from a wide variety of brands. The
Sustainable development is commonly defined as
survival of these organizations depends, to an increasing
“development that meets the needs of the present without
extent, on the quality of the products and services provided.
compromising the ability of future generations to meet their
The need for quality is also present in the software industry,
own needs” [9]. According to the United Nations World
which has consequently become concerned about software
Commission on Environment and Development [17],
quality. This has led to the appearance of the ISO/IEC 25000
sustainable development needs to satisfy the requirements of
series of standards, representing the evolution of the ISO/IEC
the three dimensions of society, economy, and the
9126 and the ISO/IEC 14598 series. This family is divided
environment.
into five divisions, one of which, the ISO/IEC 2501n (and
Sustainability has recently become more and more
more concretely the 25010), presents various software quality
important to businesses. A business that fails to have a
models.
sustainable development as one of its top priorities could
receive considerable public criticism and subsequently lose Economic sustainability is related to aspects of the
market legitimacy [6]. The authors of [6] state that, according software business, but not to its development;
to a global IBM survey in 2008, 47% of organizations have Environmental sustainability deals with technical
started to redesign their business models on the basis of aspects of software development.
sustainability, treating sustainable development as a new source Software product development mainly affects the
of innovation, an opportunity for cutting costs, and an environment via the consumption of resources that occurs
mechanism by which to gain competitive advantages; all of during its use and production. The most direct (and obvious)
which can be summarized under the concept of “strategic impact of a software product is energy consumption, but other
sustainability”, introduced by [15] . resources may also have a negative impact on the software’s
As noted by [6], technology is doubly important to pursue sustainability. The use of a processor, increased needs of
strategic sustainability. On the one hand because it helps memory and disk storage, network utilization and bandwidth,
organizations tackle environmental issues (using web the potential relocation of software production and use, among
conferences, repositories, etc.) and on the other because others, are also elements to take into account.
technology is often responsible for major environmental We believe that it is of prime importance to pay the
degradation (amounts of energy consumed by the engineering necessary attention to the environmental dimension of
processes needed to manufacture products). This mixed role sustainability in the software context, which we term as green
puts technology organizations under tremendous conflicting software or software greenability. figure 1 provides a graphic
pressures. Internally they are pressed to transform existing representation of our research focus.
engineering processes to make them more environmentally
friendly, while externally they are expected to design new
products that improve the sustainability of society.
III. SOFTWARE SUSTAINABILITY
Although there have been initiatives related to Green IT, Software Sustainability
efforts in the Green software area are still in early stages.
Software development should not remain indifferent to the
need to construct software products that contribute towards
sustainability during both their creation and use. Software is the
Software Software
core of any IT technology, and the way by which it is
Social Economic
developed can greatly influence the activities that need this
software to be accomplished, such as the functions provided or Sustainability Sustainability
how the IT infrastructure is used.
However, while sustainability is a standardized practice in a
number of engineering disciplines, efforts in software
engineering are recent and still immature [14], and the way to Software
achieve sustainable software is mainly by improving its power Environmental
consumption [2]. However, this is a very restrictive Sustainability
interpretation of what software sustainability is. (Software
Sustainable software is a “software whose direct and Greenability)
indirect negative impacts on economy, society, human beings
and the environment that result from the development,
deployment and usage of the software are minimal and/or have
a positive effect on sustainable development” [5].
This idea can be extended to cover the whole software Fig. 1. Greenability as environmental sustainability
development process. Sustainable Software Engineering can
thus be referred as “the art of defining and developing software Furthermore, according to [1], there are three
products in such a way that the negative and positive impacts environmental impact levels of ICTs:
on sustainability that result and/or are expected to result from 1st order: direct environmental effects of production
the software product over its whole lifecycle are continuously and use of ICTs;
assessed, documented, and optimized” [4]. 2nd order: indirect effects of using the system
The UN identifies three dimensions for sustainable (resource and energy conservation owing to
development – social sustainability, economical sustainability optimization or substitution of product materials);
and environmental sustainability [17]. We relate them to 3rd order: indirect, long term, effects of using the
software as follows: system (lifestyle changes that may impact on the
Social sustainability is related to software use (by environment).
whom, how and under what circumstances software We believe that, during development, the 1st level impact
may be used); can be taken into account more easily because it is related to
the direct effects of the software. The other levels will depend This international standard defines three quality models.
on how the software will be used, environmental aspects, and Two of them, the product quality model and the quality in use
other external factors out of the developer’s control. Therefore, model, are related to the product, while the third is related to
our work focus on the 1st level impacts, which will to some the quality of data. Our work is focused on the first two.
extent have an influence on the second level (figure 2).
Dimensions for sustainable development (UN, 1987)
Social Sustainability
Economic sustainability
Environmental sustainability
1st level 2nd level 3rd level
Impact impact impact
Environmental impact levels of ICTs (Berkhout and Herin, 2001)
Fig. 2. TIC impact levels and the greenability
Fig. 4. Targets of quality model [10]
IV. SOFTWARE QUALITY AND GREENABILITY
When a software product is developed it is necessary to The software product quality model is composed of eight
specify the requirements that the product should satisfy. characteristics, which are further subdivided into sub-
Software requirements can be classified into functional and characteristics that can be measured internally or externally.
non-functional requirements. As is stated in the standard, “the product quality model is
The former should define the fundamental actions that must useful for specifying requirements, establishing measures, and
take place in the software in accepting and processing the performing quality evaluations” [10]. The quality
inputs and in processing and generating the outputs [8]. The characteristics defined can be used as a checklist in order to
functional requirements are therefore related to the “What” of a ensure a comprehensive treatment of quality requirements, thus
software product. providing a basis that can be used to estimate the consequent
Non-functional requirements can be seen as requirements effort and activities that will be needed during systems
that constrain or set some quality attributes upon functionalities development. The characteristics in the product quality model
[7]. This means that non-functional quality requirements can be are intended to be used as a set when specifying or evaluating
seen as the “How” of a software product. software product quality”.
Bearing in mind that software greenability is a means to The system quality in use model is composed of five
improve a software product, we believe that it must be part of characteristics, which are further subdivided into sub-
quality and, therefore, it is related to non-functional characteristics that can be measured when a product is used in a
requirements. The first step should therefore be to include realistic context. These characteristics are thus a starting point
greenability in software quality. for requirements, and can be used to measure the impact of the
According to ISO/IEC 25010 (figure 3), process quality quality of the system in use and maintenance.
influences product quality, which in turn influences quality in The software product quality characteristics can be used to
use. On the other hand, quality in use depends on the software specify and evaluate detailed characteristics of the software
product, which depends on process quality. product that are prerequisites for achieving desired levels of
quality in use.
A. Software Product Greenability
In [3], we have proposed the inclusion of sustainability in
the software product model of the ISO/IEC 25010 standard
[10]. However, naming the characteristic “sustainability” was
not ideal because, as explained in section III, greenability deals
Fig. 3. Software quality life cycle [10] with the technical aspects of the sustainability (see figure 1).
Furthermore, the proposal did not consider capacity
The quality of a system is defined as the degree to which optimization as part of this greenability. We therefore refine
the system satisfies the stated and implied needs of its various that proposal as follows:
stakeholders, and thus provides value. These stated and implied The proposal includes four sub-characteristics for the
needs are represented in the SQuaRE series of standards in product greenability characteristic (figure 5):
various models (figure 4).
Energy efficiency. Degree of effectiveness and Note that this is not a linear process, but an iterative one, as
efficiency with which a software product consumes the completion of one action may lead to the review of the
energy when performing its functions. previous ones.
Resource optimization. Degree to which the resources
expended by a software product, when performing its A. Working with the sub-characteristics
functions, are used in an optimal manner. As in the
standard, the authors consider that resources can We identified the sub-characteristics in two steps. First, we
include: other software products, the software and studied the model characteristics in order to determine which
hardware configuration of the system, and materials existing characteristics would affect greenability. Secondly, we
(such as print paper, storage media). considered adding new sub-characteristics not derived from the
Capacity optimization. Degree to which the maximum previous step. We have done this by using the definition of the
limits of a product or system parameter meet characteristics provided in the standard trying to determine its
requirements in an optimal manner, allocating only influence on the greenability in use. We have consequently
obtained the following sub-characteristics:
those which are necessary.
Perdurability. Degree to which a software product can Efficiency Optimization. Optimization of resources
be used over a long period, being, therefore, easy to expended in relation to the accuracy and completeness
with which users achieve goals. Relevant resources can
modify, adapt and reuse.
include time consumption, software resources, etc.
The next step in the process of integrating greenability into User’s environmental perception. Degree to which
the software quality should be to include it in the quality in use users are satisfied with their perception of the
model presented in ISO/IEC 25010. consequences that the use of software will have on the
Therefore, software product greenability can be defined as: environment.
Degree to which a product lasts over time, optimizing the Minimization of environmental effects. Degree to
parameters, the amounts of energy and the resources used. which a product or system reduces the effects on the
environment in the intended contexts of use.
B. Defining new characteristics
New characteristics are formally defined and refined,
considering the sub-characteristics identified above. In this
case, only one new characteristic has been identified:
Greenability in use. Degree to which a software product
Fig. 5. Software product greenability
can be used by optimizing its efficiency, by minimizing
environmental effects and by improving the environmental user
Greenability, however, is not only relevant to the software
perception.
product. As shown in figure 3, software product quality
influences quality in use and quality in use, depends on the
Because of the two first steps, we have obtained the new
software product quality. So, it is also necessary to study the
characteristic and sub-characteristics shown in figure 6.
inclusion for the greenability in the quality in use model.
V. ADDING GREENABILITY TO THE QUALITY IN USE MODEL
Greenability in
The process of including a new characteristic in the use
standard requires a set of actions:
A. Working with the sub-characteristics: The
identification and definition of the sub-characteristics User's Minimization of
Efficiency
is carried out. Environmental environmental
Optimization
perception effects
B. Defining the new characteristics: new characteristics
formally defined/refined, considering the sub- Fig. 6. Quality in use Greenability
characteristics.
C. Reviewing quality in use characteristics: the model is
reviewed in order to check whether it is affected by the C. Reviewing quality in use characteristics
inclusion of the new characteristic.
It is now necessary to review the definitions of the other
D. Redefining quality in use: the quality in use definition
quality in use characteristics that could be affected by the
is reviewed in order to include the new characteristic in
inclusion of the new one. The following definitions have
it.
therefore been redefined:
Context coverage. Degree to which a product or system
can be used with effectiveness, efficiency, freedom from risk,
greenability and satisfaction in both specified contexts of use this option and it is necessary to generate an expense
and in contexts beyond those initially explicitly identified report for each department. This implies that in order
Context completeness. Degree to which a product or for the user to perform this task, s/he must use more
system can be used with effectiveness, efficiency, freedom from resources.
risk, greenability and satisfaction in all the specified contexts Positive. The application provides accurate information
of use about the enterprise’s light consumption expenditure.
Flexibility. Degree to which a product or system can be
used with effectiveness, efficiency, freedom from risk, Second level impact
greenability and satisfaction in contexts beyond those initially Negative: The data recovery processes are slower
specified in the requirements because data are scattered in several reports.
D. Redefining quality in use Positive: Reports are stored in digital format,
signifying that reports on paper are not necessary (e-
Finally, the last step is to review the quality in use
materialization).
definition in such a manner that it takes into account the new
added characteristic.
Third level impact
Quality in use is the degree to which a product or system A growth in IT-related services supposes a structural
can be used by specific users to meet their needs in order to change and a new way of using resources at company
achieve specific goals with efficiency, freedom from risk, level.
greenability and satisfaction in specific contexts of use. As will be noted, the development of a green software
product impacts directly on the first level but also influences
the second and may even reach the third.
The result of this process is the new quality in use model
shown in figure 7. Having defined the product quality (PQ) and the quality in
use (QiU) models, the next step is to discover how to link them.
Quality in use VII. LINKING PRODUCT QUALITY AND QUALITY IN USE
By incorporating greenability to the software product
Freedom from Context
Effectiveness Efficiency Satisfaction
risk coverage
Greenability quality and to the quality in use models of the ISO/IEC 25010
standard, we have created the foundations for assessing and
Economic risk Context Efficiency
Usefulness
mitigation completeness Optimization achieving greenability in software. Therefore, we must
consider the greenability of both the software product itself and
Health and User's
Pleasure safety risk Flexibility Environmental of the software in use. However, as discussed in [12], the
migitation perception
ISO/IEC 25010 standard (figure 3) states that there is a
Minimization of
Comfort
Environmental
risk mitigation
environmental relationship between the product quality and the quality in use,
effects
but it does not say how to make this connection.
Trust
In fact, in the software quality field, most efforts have been
made as regards product quality and it is difficult to find works
on quality in use. The emphasis has principally been placed on
Fig. 7. Complete Quality in use model assessing the quality of software products because they are
VI. STUDYING THE IMPACT LEVELS OF ICTS WITH REGARD TO more precisely defined in literature, can be more easily
identified (and are thus easier to evaluate), and also because of
GREENABILITY.
the “dependency” of QiU on product quality.
As previously indicated our work is focused on minimizing This relationship between the product and the quality in-use
the first level impact of software, although it may also affect of the software product would appear to be based on the
the second and third levels (see Section II and figure 2). It must assumption than having a product with high quality will
be noted that effects at the third level are indirect and more guarantee a product with a high QiU. However, this is not
difficult to consider during the software development. necessarily true in many situations. The fact that a product has
In this section, we wish to provide an example of the the best quality does not necessarily guarantee that the product
relationship between a greenability in use sub-characteristics will fulfill the user’s needs in its context of use, especially
and the ICT impact levels. when the overall quality (as perceived by the end-user) is
A. Sub-characteristic: Efficiency Optimization composed of many conflicting factors. A Ferrari is not the best
car to go to work in if you are a social worker in a deprived
Domain application: Application with which to generate
suburb in the outskirts of New York.
reports
Our focus on the quality assessment is exactly the opposite:
First level impact
concentrating on the quality in use as the driving factor to
Negative. The user wishes to obtain an annual report of consider when designing a software product, or when selecting
the company’s total expenditure grouped by the product that best fits a user’s needs. There are several
department. However, the application does not provide reasons why we feel the need to challenge the traditional
approach used to evaluate the quality of a software product. The trained network can additionally be used to make
Firstly, not all the product quality characteristics of a software inferences about the values of the variables in the network.
product have the same influence on its QiU. This, together with Bayesian propagation algorithms use probability theory to
the false assumption regarding the direct dependency between make such inferences, using the information available (usually
the product quality and the QiU mentioned above, frequently a set of observations or evidences). Such inferences can be
forces some of the product aspects (which are non-critical for abductive, and if we wish to determine the product quality sub-
the end user) to be over-specified for the sake of ensuring a characteristics we must consider guaranteeing a required level
certain level of QiU. This unnecessarily increases costs, of QiU (the cause that best explains the evidence); or
development efforts and, resource usage, without a direct effect predictive, if we wish to determine the probability of obtaining
on the advantages that the end-user perceives. certain results in the future. All the variables in the network can
We therefore focus on the QiU, and we analyze the therefore be used as either a source of information or an object
relationship between the product quality and the QiU of a of prediction, depending on the evidence available and on the
software product in the opposite direction to that which has goal of the diagnostic process.
traditionally occurred. This means applying a ‘backward’
B. Modelling the Bayesian Network
analysis (we start with a given level of QiU and we wish to
determine the minimum level of product quality that will In this point, we show how to prepare a Bayesian Network
guarantee such a desired quality in use), as opposed to the in order to link product quality and quality in use. It should be
traditional “forward” analysis (by which we attempt to noted that this is a general example as regards how to use this
determine the level of QiU of a software product, given a approach, and that it must be tailored to specific contexts.
measured level of product quality). In order to obtain a (good) Our working hypothesis is that the PQ has an influence on
level of quality in use, the goal would in fact be to be able to the QiU and that this influence can be modeled and studied by
select the reduced set of really relevant product quality sub- using a BN, in such a way that we can conduct a backward
characteristics that ensure the required level of quality. analysis of the required level of PQ to ensure a given level of
Focusing solely on them will avoid superfluous costs or QiU. We have successfully applied this approach previously
irrelevant features which may unnecessarily increase the final [12], and other authors have also used the BN for the
impact on the environment and also the price of the product. assessment of software quality [13].
In order to model the Bayesian Network, it is first necessary
A. Use of Bayesian Networks to identify the relationships between the PQ characteristics and
In order to determine the relationship between the quality in the Quality in use. The relationship between PQ and QiU can
use (QiU) and the product quality (PQ), we need statistical be modeled by determining the characteristics of the former
methods and tools that can carry out backward analyses. that affect the characteristics of the latter. As our present focus
However, the commonly used linear regression (LR) or is on greenability, we shall focus the process solely on the
principal component analysis (PCA) are not useful here greenability in use characteristic (although the same process
because they conduct forward analysis and need initial can be applied to the other QiU characteristics).
numerical information at data level, which is in many cases It is then necessary to identify the PQ characteristics that
difficult to obtain. However, Bayesian Belief Networks (or, have a significant influence on the greenability in use sub-
simply, Bayesian Networks, BNs) can be very useful. A BN is characteristics. This is done by using the definitions provided
a directed acyclic graph, whose nodes are the uncertain in the standard along with those defined in this paper for the
variables and whose edges are the casual or influential links new sub-characteristics related to greenability. This process is
between variables. A Conditional Probability Table (CPT) is made by a set of experts in quality. Firstly, they established the
associated with each node in order to denote such causal relationship among the characteristics independently. Next,
influence [11]. they made a meeting to interchange opinions and agree on
To define a BN it is necessary to: (1) provide the set of values. Table 1 shows these relationships by employing a
random variables (nodes) and the set of relationships (causal matrix, in which the “X” indicates a relationship between these
influence) among those variables; (2) build a graph structure characteristics.
with them; and (3) define conditional probability tables The next step is to determine the relationship between the
associated with the nodes. These tables determine the weight sub-characteristics of the QP characteristics and the sub-
(strength) of the links of the graph and are used to calculate the characteristics of the greenability (using the information in
probability distribution of each node in the BN. Table I as a starting point). This step is necessary because if it
The use of BNs thus allows us to model the different were not taken we would be assuming that, for example, all the
relationships among the characteristics and sub-characteristics reliability sub-characteristics have the same influence on the
of the product quality and the QiU, in addition to the degree of efficiency optimization and this might not be true (it could
dependence or influence among them. This signifies that it is perhaps be more closely related to fault tolerance than to
necessary to define the structure and conditional probability maturity).
tables in which the uncertainty relationships among the BN We have chosen only one of the greenability in use sub-
nodes (characteristics) that we wish to build are reflected. Once characteristics in order to continue with the example. More
the network has been defined, it is necessary to train it using a specifically, and based on the results shown in Table I, we
set of controlled experiments, so that it “learns”. decided to work with the environmental user perception
because it is the one with most interactions and because it models and the construction rules for Bayesian Networks.
appears to be the closest to the user. figure 8 shows the results.
As can be observed, there is a node for each characteristic
Greenability in Use and sub-characteristic, and arcs represent the relationships
between the nodes (there is one arc for each X in Table II and
Efficiency
User’s Minimization of another with which to connect a PQ characteristic with its sub-
environmental environmental
optimization
perception. effects.
characteristics).
Functional
X X User’s User’s
Suitability
environmental environmental
Performance perception. perception
X X X
Efficiency
Compatibility X X X Adaptability X Maturity X
Product quality
Usability X X Portability Installability Availability
Reliability
Replaceability X Fault tolerance X
Reliability X
Co-existence Recoverability X
Security Compatibility
Interoperability X Modularity X
Maintainability X X X Appropriateness
X Reusability X
recognisability
Portability X X
Operability Analysability
Maintainability
Greenability X X X User error
Modifiability X
protection
6 8 5 Usability
User interface
Testeability
aesthetics
Energy
Learnability X X
efficiency
TABLE I. RELATIONSHIPS BETWEEN CHARACTERISTICS OF SOFTWARE Accessibility Greenability Resource
X
PRODUCT QUALITY AND GREENABILITY IN USE optimization
Functional
Perdurability X
Completeness
We have therefore identified the relationships shown in Functional Functional
Table II. The “X” again indicates a relationship between these Suitability Correctness Performance
Time- behaviour X
Efficiency
characteristics. Functional
Resource
utilization
X
X
Appropriateness
The level of influence (indicated by the relationships in Table I Capacity X
and Table II) may vary between different application domains.
The relationships shown in both tables might then need to be
TABLE II. RELATIONSHIPS BETWEEN PRODUCT QUALITY SUB-
tailored to other domains, but the method indicated in this CHARACTERISTICS AND GREENABILITY IN USE SUB-
paper is still valid. CHARACTERISTIC
It is necessary to use the information provided in Table II to
model the Bayesian Network that reflects the relationships
identified, taking into account the hierarchical structure of the
Resource
utilisation
Time behaviour Capacity Approp.
Learnability
Recogn.
Performance
Recoverability
Adaptability Replaceability Efficiency Usability
Maturity Fault-tolerance
Appropriateness
Portability
Reliability
Interoperability
Energy
Analysability efficiency
Resource
Modularity Modifiability Perdurability
optimisation
Maintainability Greenability
User’s environmental
perception
Fig. 8. BN for environmental user perception
Although this BN reflects the relationships identified, it among the PQ characteristics that are conceptually related. The
produces a very high number of entries on the final node (that BN obtained (figure 9) drastically reduces the number of
of the environmental user perception). The definition of the entries in the probability tables. The BN obtained could easily
probability tables is therefore very laborious and cumbersome. be used to create three individual BNs (one for each of the sub-
One practice that is commonly used to simplify the trees that comprise the BN), work independently with them and
relationships in BNs is based on the introduction of synthetic then combine them to form the complete Bayesian Network.
nodes. In this case, we decided to introduce a synthetic node
Approp.
Learnability
Recogn.
Adaptability Replaceability
Recoverability
UsabilityUEP
Maturity Fault-tolerance Resource
PortabilityUEP utilisation
Analysability Time behaviour Capacity
ReliabilityUEP
Modularity Modifiability
Performance
EfficiencyUEP
MaintainabilityUEP Interoperability Energy Capacity
UsageUEP efficiency Optimization
Resource
Resource optimisation Perdurability
AspectsUEP
Functional
Functionality
Appropriateness UEP GreenabilityUEP
User’s environmental
perception
Fig. 9. The BN from Fig 7, with synthetic nodes
calculated for the product; this will preferably be automatic,
C. Adapt the Bayesian Network
though that is not always possible. These measurements will be
In the previous point, we have shown an example of how to the ones which will serve as input to the external nodes of the
use Bayesian Networks to determine the influence of the network; their values should be changed into valid inputs to the
greenability in use on the product quality. However, it is network. The values will be propagated though the BN, via the
necessary to define the rest of BNs and adapt them to the nodes and by applying the probability tables, until the lower
specific context to which it is to be applied. node is reached (the one about quality in use, or some of its
In order to carry out this adaptation, we must ensure that all characteristics)
the characteristics of the standard are applicable to this context
and that no further characteristics are going to be needed. In
addition, it should be determined whether to include new D. Use of the Bayesian Networks
characteristics of the context by studying the state of the art,
looking for other proposals, consulting experts, etc. Once the BN’s have been adapted to the specific context,
After these actions have been taken, we will be able to they are ready to be used. These BN’s can be used to carry out
build the structure of the Bayesian network. a forward or backward analysis.
The next step is to create the probability tables. The In the forward analysis, we can determine the quality in use
influences of some given characteristics will obviously depend of a product once it has been created. In order to do we should
on the domain. That means it is vital to create tables to reflect define measurements for the external nodes that make up the
the specific reality of a particular domain. input to the network.
To do that, we must carry out experiments or surveys that In the backward analysis, we can determine the minimum
allow us to obtain a series of data that serve as input to the values of external quality that the product needs to reach a
network validation process. This is part of our future work in desired level of quality in use. In that way, we can ascertain
order to finish the Bayesian network and use it for working on what values (for the measurements defined for each
the greenability assessment of a software product. characteristic) the product should have for the input nodes.
The final step in being able to use this network will be the
definition of specific measurements for the software product
we wish to measure. These measurements should be able to be
VIII. CONCLUSIONS AND FUTURE WORK Sustainability, Resilience in 9th IFIP TC 9 International
Greenability is a means to improve the software product Conference, HCC9 2010 and 1st IFIP TC 11 International
and should therefore be integrated into quality models, such as Conference, CIP 2010, Held as Part of WCC 2010. 2010.
the ISO/IEC25010 standard. By doing so, greenability can be pp. 248–259. Available from:
assessed and achieved during the software development http://dx.doi.org/10.1007/978-3-642-15479-9_24.
process, just like other software qualities. [6] Du, W.D., S.L. Pan, and M. Zuo, "How to Balance
In this paper, we have proposed a extension of the quality Sustainability and Profitability in Technology
in use model of the ISO/IEC25010 standard, includes a new Organizations: An Ambidextrous Perspective.". IEEE
characteristic: greenability. This is composed of three sub- Transactions on engineering management. Vol. 60(2): pp.
characteristics: Efficiency Optimization, User’s environmental 366-385 10.1109/TEM.2012.2206113, 2013
perception, and Minimization of environmental effects. [7] Glinz, M., On non-functional requirements. "International
Moreover, we consider that there is a direct influence Conference on Requirements Engineering": pp. 21-26,
between the product quality (which includes a greenability 2007
characteristic) and the quality in use (which already integrates [8] IEEE 830, IEEE Recommended Practice for Software
greenability). We have therefore proposed a Bayesian Network Requirements Specifications. . 1998
that shows the relationships between both. As a future work we [9] ISO 26000 Guidance on Social Responsibility. 2010 Available
plan to work on this Bayesian Network, apply it to a specific from: https://www.iso.org/obp/ui/#iso:std:iso:26000:ed-
domain and construct the probability tables in order to assess 1:v1:en
the greenability level of a given software product. We also plan
[10] ISO/IEC 25010, Systems and software engineering - Software
to use Bayesian Networks for the greenability evaluation by
product Quality Requirements and Evaluation (SQuaRE) -
means of measures and indicators. This is therefore another of
Software product quality and system quality in use models.
our future works.
2010
We also wish to continue with our work by studying the
[11] Jensen, F.V., Bayesian Networks and Decisions Graphs.:
other aspects of sustainability, i.e., the economic and mainly
the social aspects to which we believe special attention should Springer-Verlag, 2001.
be paid in order to indicate and mitigate some labor situations [12] Moraga, M.A., M.F. Bertoa, C. Morcillo, C. Calero, and A.
that currently occur in the software industry and that should be Vallecillo Evaluating Quality-in-Use Using Bayesian
rejected immediately. Networks. in 12th ECOOP Workshop Quantitative
Approaches on Object Oriented Software Engineering
ACKNOWLEDGMENT (QAOOSE 2008). 2008.
This work has been funded by the GEODAS-BC [13] Neil, M., P. Krause, and N.E. Fenton, Software Quality
project (Ministerio de Economía y Competitividad and Fondo Prediction Using Bayesian Networks: Software
Europeo de Desarrollo Regional FEDER, TIN2012-37493- Engineering with Computational Intelligence, Chapter 6,
C03-01). Kluwer, 2003.
[14] Penzenstadler, B., A. Raturi, D. Richardson, C. Calero, H.
REFERENCES Femmer, and X. Franch Systematic Mapping Study on
[1] Berkhout, F. and J. Hertin, Impact of Information and Software Engineering for Sustainability (SE4S) in 18th Intl.
Communication Technologies on Environmetal Conf on Evaluation and Assessment in Software
Sustainability: speculations and evidence. 2001 Engineering. 2014.
[2] Calero, C., M.F. Bertoa, and M.A. Moraga A Systematic [15] Sroufe, R. and J. Sarkis, Strategic Sustainability: The State of
Literature Review for Software Sustainability Measures in theArt in Corporate Environmental Management Systems:
GREENS 2013 : Second International Workshop on Green Sheffield, UK: Greenleef Publishing, 2007.
and Sustainable Software. 2013. pp. 46-53. [16] SWEBOK, "SWEBOK V3.0. Guide to the Software Engineering
[3] Calero, C., M.A. Moraga, and M.F. Bertoa Towards a software Body of Knowledge". Bourque, P. and Fairley, R.E. (eds.),
product sustainability model in Workshop on Sustainable NJ. IEEE Computer Society. . Vol., 2014
Software for Science: Practice and Experiences 2013. [17] United Nations World Commission on Environment and
arXiv:1309.1640 2013. Development, Report of the World Commission on
[4] Dick, M. and S. Naumann Enhancing software engineering Environment and Development: Our Common Future. In
processes towards sustainable software product design in United Nations Conference on Environment and
EnviroInfo 2010, Integration of Environmental Information Development,. 1987
in Europe, Proceedings of the 24th International
Conference on Informatics for Environmental Protection.
2010. pp. 706–715.
[5] Dick, M., S. Naumann, and N. Kuhn A Model and Selected
Instances of Green and Sustainable Software. What Kind of
Information Society? Governance, Virtuality, Surveillance,