-This paper introduces a framework for software sustainability profiling. The goal of the framework is to analyse sustainability requirements for long-living software systems, focusing on usability and readability of the sustainability profiles. To achieve this goal, we apply a quantitate approach such as fuzzy rating scale-based questionnaires to rank the sustainability requirements, and the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) to analyse the results of questionnaires and to provide a basis for system profiling. The core profiling elements provided by our framework are (1) a sustainability five-star rating, (2) visualisation of the five sustainability dimensions as a pentagon graph detailing combination for individual, social, technical, economic and environmental dimensions, and (3) a bar graph of overall sustainability level for each requirement. To ensure sustainability, the proposed profiling framework covers the five dimensions of sustainability to quantify the sustainability of any software system not only during the requirement gathering phase but also during maintenance phase of software system lifecycle.
I. INTRODUCTION
Addressing the impacts of software systems on
sustainability is a first-class quality concern beside usability, safety and
security [
The analysis of system sustainability has to be initiated on
the requirements engineering (RE) phase [
This question is especially important for long-living systems, where the stakeholders requirements and preferences might change over the time the system is in use. For example, a system that can be seen as sustainable today, might be rated as environmentally unsustainable in few years, while new techniques to increase environmental sustainability are developed. To solve this problem, we require an easy-to-use profiling framework based on quantitate approaches that would allow to analyse the up-to-date system sustainability profiles, based on system characteristics and the up-to-date ratings (quotations) of the corresponding requirements. Usability and readability of the approach is crucial to make it applicable for real software development processes, as the quotation process and the generated profiles have to be easy-to-use by all stakeholders.
Contributions: To ensure the sustainability of long-living
software systems over their eitire live-cycle, we propose a
framework for sustainability profiling. The framework allows
to analyse sustainability requirements for long-living software
systems. The up-to-date profiles could be generated both
during the RE and the maintenance phase of the software
system lifecycle. The framework workflow is presented in
Figure 1. First of all, stakeholders are assigned to a group
to rate requirements from the different perspective of
sustainability dimensions (individual, social, technical, economic
and environmental). Then, a fuzzy rating scale is used to
avoid imprecision for answering quantitative questionnaires
[
Outline: The rest of the paper is organised as follows. In Section II we discuss the background and related work. Section III introduces our framework for software sustainability profiling. Section IV introduces an example scenario to show how the framework can be used to profile software systems. Section V summarises the core contributions of our work.
In this section we discuss the research directions and
approaches that provide a background for our framework: RE
for sustainable systems, the idea of the sustainability profiling,
quantitative approaches, approaches using the fuzzy rating
scale, and the TOPSIS framework for requirements analysis.
We selected TOPSIS for our sustainability profiling
framework, as this technique has been successfully used for
prioritising requirements and solving conflict among non-functional
requirements, cf. [
The RE phase of software development focuses on
discovering, developing, tracing, analysing, qualifying, communicating
and managing system requirements, cf. e.g., [
Penzenstadler [
Sustainability in software has various dimensions. Goodland
[
In our framework, we analyse the system sustainability using the five dimensions:
Individual sustainability: Individual needs should be protected and supported with dignity and in a way that developments should improve the quality of human life and not threaten human beings; Social sustainability: Relationships of people within society should be equitable, diverse, connected and democratic; Technical sustainability: Technology must cope with changes and evolution in a fair manner, respecting natural resources; Environmental sustainability: Natural resources have to be protected from human needs and wastes; and Economic sustainability: A positive economic value and capital should be ensured and preserved.
Sustainability profiling has been used mostly for software
energy and data centre consumption, as well as in cities
and urban settlements. James [
Gmach et al. [
Step 3 Determining a support response to be considered as compatible to some extent; and Step 4 Creating a trapezoidal fuzzy number from the two intervals, which are linearly interpolated, as T ra(a; b; c; d), where 0 a b c d 1. of any software system, starting from the requirements phase and continuing over the phase of maintenance.
Quantitative approaches are used to analyse data and to
measure qualities in software engineering [
The quantitate approaches can be applied to several types of
data, and the type of data to analyse might influence the choice
of the approach. Tullis and Albert [
Nominal data is categorised or classification data, which it is not in any particular order, e.g., gender or hair colour; Ordinal data is ordered classified data, but the differences between them are not meaningful, e.g., product and movie ratings; Interval data is classified data where the difference between two data items is meaningful, but without natural zero points, e.g., temperature units; Ratio data is interval data with absolute zero, e.g., weight and height.
To analyse sustainability requirements, we will create from the provided by stakeholders ranking the corresponding ratio data. This transformation will be done using TOPSIS, cf. Section II-E. The ratio data will be then further explored to build the system profile.
A fuzzy rating scale (FRS) allows the capturing of the
diversity of individual responses in questionnaires, also
avoiding imprecision while rating a questionnaire [
To implement an FRS, we adopt the fuzzy rating scale
method proposed by Lubiano et al. [
Fig. 2. Fuzzy Rating Scale for Sustainability Profiling
Technique for Order of Preference by Similarity to Ideal
Solution (TOPSIS) is an effective technique to evaluate
sustainability requirements which change over time is utilising.
TOPSIS is one of the multiple criteria decision analysis
approaches to identify the best alternative that is nearest to
an ideal solution and farthest from negative ideal solution
[
The following is the stepwise procedure of TOPSIS
according to Behzadian [
xij qPm i=1 xi2j ; f or i = 1; ; m; j = 1; ; n (1) Step 2 Construct the weighted normalised decision matrix vij
Similarly, the separation from negative ideal is Step 5 Calculate the relative closeness to the ideal solution Ci
S0 Ci = (S + S0 ) ; 0 < Ci < 1; i = f1; ; mg (7) Ci = 1 Ci = 0 if Ai solution has the best condition, if Ai solution has the worst condition.
III. FRAMEWORK FOR SUSTAINABILITY PROFILING The general idea of the framework workflow is presented in Figure 1. To measure the sustainability aspects of the requirements, we adopted the FRS approach. Requirements are rated against sustainability dimensions, which gives an input to the TOPSIS procedure. The provided by TOPSIS results will create a basis for sustainability profiling: using these results, our framework determines (1) the sustainable of each system requirement, (2) sustainability of the software system as whole. This will be presented in a five-star rating within each level of sustainability dimensions and the overall sustainability of each requirement. The analytical approach consists of the following five steps, cf. also Figure 1.
Requirements engineers should assign stakeholders to one of the three stakeholder groups having end-users, administrators, and developers and providers groups. For instance, in eLearning systems the learner and instructor are in the end-users group while ITs support could be assigned to the administrator group. (2) ; vng
(3) ; vn0g (4) (5) (6)
The framework will generate a questionnaire including related questions (instructions) for each requirement with regard to the sustainability dimensions and stakeholders groups. Thus, for each requirement k questions will be created, where 1 k 5. Each question should present a single sustainability dimension perspective, which is covered by the requirement, and have a form
where X is belongs to the set findividual; social; technical; environmental; economicg. The generated questionnaire can be further revised and adapted by both requirements engineers and sustainability experts, before continuing with the next step.
For example, requirement R1 has to have five questions, covering each dimension of the sustainability.
Each stakeholder has to answer allotted question from vary views of certain sustainability dimension by using the FRS. For example, stakeholders, who are in the learners and instructors group, will answer two questions for each requirement: from the individual and from the social sustainability point of view. and another time for the social sustainability. Each answer, also, will be in a form of trapezoidal fuzzy number from the two intervals as T ra(a; b; c; d), where 0 a b c d 1.
After all stakeholders answered the questionnaire, the results of the FRS approach become inputs for TOPSIS. The data will be normalised and weighted according to Equations 1 and 2, and after that the steps 3, 4 and 5 of TOPSIS need to be applied twice:
First round: Applying requirements as criteria to determine overall sustainability within the separation of requirements’ impacts for each requirement; and Second round: Applying sustainability dimensions as criteria to analysis each dimension within all requirements and overall sustainability rating for the software.
The result of TOPSIS analysis including two rounds helps to generate software sustainability profiling which is visualised representing the result. The profiling includes:
Sustainability five-star rating Presenting the average of P C in the both rounds of sustainability dimensions and requirements; Five sustainability dimensions Illustrating each dimension level combined in pentagon or bar graph (optional) for the software having all rated requirements; and Bar graph Showing an overall sustainability for each requirement.
An example of a sustainability profile for a software system, which is created using the proposed framework, is presented in the next section (cf. Figure 5).
Considering a different information in the profiling, we simplify and visualise the result by creating a key chart with five categories as shown in Table I. This key chart includes numeric variables in percentage, colour codes for visualisation, and linguistic variables as a description.
Figure 3 demonstrates how the proposed framework can be
uses during the RE activities (we follow the definition of the
RE activities introduced by [
Requirements elicitation is the practice of understanding and determining stakeholders’ needs and constraints. To rate the sustainability requirements using the proposed framework, at this phase two actions are necessary: (A) the stakeholders have to be assigned, (B) the questionnaires have to be generated.
However, taking into account the long-living nature of the system, re-iteration of these steps might be necessary on the management phase, to ensure the sustainability over the software system lifecycle: (A0) new stakeholders can be assigned, (B0) the questionnaires can be updated. Requirements analysis is the practice of refining stakeholders’ needs and constraints by defining the process, data and object of the required system.
On this phase, we conduct the following steps of our framework: (C) the stakeholders rate the requirements, (D) the sustainability of the system is analyses using TOPSIS, (E) the sustainability profile is generated.
To ensure longevity of the system, these steps also can be repeated during the management phase.
Requirements specification is the practice of writing down stakeholders’ needs and constraints, and this documentation should be unambiguous, complete, correct, understandable, consistent, concise, and feasible.
The sustainability profile could be seen as one of the input to the specification phase.
Requirements validation is the practice of checking that the specification captures users’ needs and constraints. The proposed framework does not cover the validation activities, which might be one of the future work directions.
Requirements management is the practice of scheduling, controlling changes and tracking requirements over time. In the case of long-living systems, the management activities are crucial to keep the software system sustainable. The steps (A) (E) have to be repeated to provide an up-to-date sustainability profile of the system.
IV. APPLICATION OF THE PROPOSED FRAMEWORK Let us discuss an example scenario with five requirements R1, . . . , R5. The aim of this scenario is to illustrate application of the proposed framework, without going into the technical details like generating of questions within real questionnaires. In this scenario, we will go through all framework steps and present the created sustainability profile as the final result.
Let us assume that the requirements will be rated by ten assigned stakeholders: four in the end-users group, three in administrators group, and three in developers and providers group.
This step is omitted in the example, as the rating activities will be simulated.
To simulate the rating activities where each stakeholder rates requirements against sustainability dimensions by answering defined questions, we generate random numbers between [0:1] (0 corresponds to a critical value, 1 corresponds to a green value) for the fuzzy rating scales. Figure 4 shows the results of application of the FRS approach to the requirement R1, from the prospective of ten assigned stakeholders.
As follows from Figure 4, Stakeholder S2, who is assigned to individual and social sustainability dimensions, rates R1 for
A Representative Rating ilcaoS CSri ti5cal 0 10 20 30 40 5.0 60 70 80 90 100( GVirberaennt ) iccoonEmCSri ti5cal 0 10 20 30 40 5.0 60 70 80 90 100( GVirberaennt ) individual perspective as T ra(0:51; 0:66; 0:856; 1:00) while social perspective as T ra(0:60; 0:66; 0:75; 0:9). We calculate fuzzy values from each fuzzy rating by mean measurement, so individual and social means of R1 for S2 are 0:754 and 0:727, respectively.
In the next step, all the FRS outputs become inputs for TOPSIS, cf. Table II. These data are normalised according to Equation 1 for the five system requirements R1, . . . , R5 within the individual, social, technical, economic and environmental dimensions of sustainability. The result of normalisation step presented in Table III.
The weighted normalisation that was constructed according to Equation 2 is showed in Table IV. Following the TOPSIS procedure, we calculate for both rounds the separation measures from positive ideal S and negative ideal solutions S0 , as well as the relative closeness C . The results are summarised in Tables V and VI. Noteworthy, we could calculate the negative impact of economic and environmental sustainability dimensions via the negative ideal solution that maximises the cost criteria and minimises the benefit criteria.
The generating software sustainability profiling is presented in Figure 5 within an overall of 49% sustainability which is the mean of P C in the two rounds (in Table V and VI). Among R1 R2 R3 R4 R5
Dimension Individual Social Technical Economic Environment Individual Social Technical Economic Environment Individual Social Technical Economic Environment Individual Social Technical Economic Environment Individual Social Technical Economic Environment
S1 0.573 0.276
idual div In
Socia l l a c i n h c e T
R 5 R 4 R 3 R 2 R 1
Environmental Overall 10 30 50 70
90 the five requirements, R1 meets the highest level as satisfactory as well as environmental dimensions. Also, individual, social, technical and economic dimensions become basic as the lowest level of software sustainability including the five requirements in this example.
In this paper, we introduced a framework for software sustainability profiling. We also presented and example scenario to provide a numerical illustration on how the framework can be applied. The framework allows to create the following profiling elements:
ranking of entire software requirements;
pentagon graph (and, optionally, also a bar graph) for all dimension levels of the entire software requirements; and 3) Bar graph for overall sustainability of each requirement.
In our framework we apply a quantitative approach to measure sustainability of the software systems. The fuzzy rating scale is utilised to overcome inexplicit choices in questionnaires and increase the usability of the framework. The TOPSIS approach for requirements analysis is used to analyse ranking within the best ideal solution and the worst ideal solution among requirements that could assist to recognise the positive and negative impacts on sustainability via maximising or minimising the benefit or cost.
In the case of long-living systems, it is crucial to keep the software system sustainable over the whole lifecycle of the system. The stakeholders requirements and preferences might change over the time the system is in use, and proposed framework allows to analyse the up-to-date system sustainability profiles, based on system characteristics and the up-to-date ratings (quotations) of the corresponding requirements.
One of the core features of the framework is readability of the sustainability profiles, which also implies the usability of the proposed framework. For example, we apply the fivestar rating to present sustainability ratings, as this rating is perceived as a common one in other areas: the five-star rating has become a standard for electricity consumption labelling in electronic appliances such as air conditioners and computer monitors, allowing an energy efficient choice by reducing energy use and emissions (i.e., to increase environmental sustainability).
We follow the traffic lights colouring schema, where critical values are marked red and green (vibrant) are marked green to increase readability and graphic visualisation. These colours and their descriptions have been used in Green IT and Sustainability Developments. To increase accessibility of our approach, we also provide another colouring option for colourchallenged people, where the red colour is replaced by blue. Finally, there are two options to present the five sustainability dimensions as a pentagon or bar graph because it might be argued that the pentagon graph could be harder to read and need more effort to analyse represented data than the bar graph, so we provide the bar graph option for representing the five sustainability dimensions.
Future work: The main direction of our future work on
the proposed framework is to develop a tool for software
sustainability profiling, allowing to perform the framework
steps within a single platform. We also would like to apply
the proposed framework to our earlier work on the analysis
of the RE aspects of ELearning systems [
The first author is supported by a scholarship from Umm Al-Qura University, Saudi Arabia.