VU University Amsterdam 1

Final Project Report: Semi-automated estimation of the cost benefits of green

ICT practices

A web-based environmental -economic impact calculator

Giuseppe Procaccianti, Hector Fernández, Patricia Lago VU University Amsterdam

This document reports on the activities and results of the project “Semi-automated estimation of the cost benefits of green ICT practices” (funded by SURFnet: Innovatieregeling 2013 Duurzame Ontwikkeling). The main result of the project is a software tool to support decision makers in (1) selecting, comparing and justifying among alternative green ICT practices; (2) analyzing the relation between the organization-specific settings and the achieved cost benefit; and (3) turning influential factors to optimal yet realistic values to achieve the highest cost-benefit ratio. This tool has been implemented as a web application and is now publicly available on the Internet. The remainder of the document is structured as follows. Section 1 describes the research activities we carried out during the project and obtained results, Section 2 introduces the functionalities of the calculator we implemented, Section 3 discusses the economic modeling aspects and Section 4 concludes the report.

1. Research Activities

This project required several research activities to be carried out. First of all, we had to elicit (from both practice and literature) a list of energy efficiency metrics to measure the economic impact of green ICT practices; subsequently, we had to formalize the obtained metrics in a machine-readable format, to make them usable in the semi-automated estimation process. In this section, all of these phases are described in detail.

1.1. Definition and extraction of green ICT metrics

Our first step was to collect and elicit green ICT metrics from both practice and literature. We performed a systematic literature review [1] that resulted in 66 green metrics, classified in 5 categories: Energy, Performance, Economic, Utilization, Pollution. Examples of relevant metrics identified in this study are: Energy Consumption, Energy Savings, Client/Server Energy Costs. Another study [2] we performed was more focused on metrics commonly used in industry. This study evaluated practices in four different focus areas: Embedded System Software, Generic Software, Data Centers/High Performance Computing and Hardware. Relevant practices identified in this study were Total Cost of Ownership (TCO), Power Usage Effectiveness (PUE), Energy from Renewable Sources. Finally, we performed a case study in partnership with a multi-national telecommunication organization, where we designed a framework called "Value of Energy" [3]. In this framework, we defined several metrics for data management in the Cloud, that can be applied to calculate the economic value of data management practices. Some examples of Value of Energy metrics are: Effective Power (power used to store valuable data in the organization), WastedPower (power used to store obsolete data that can be deleted or archived), Future Amount of Data (the expected amount of data in the future, according to data growth statistics). Starting from this previous work, we carried out additional research to analyze and build a consistent set of metrics for green ICT.

1.2. Formalization of green ICT metrics

As a result of the aforementioned studies, we elicited several metrics to describe the environmental and economic benefits of green ICT practices. Here follows a list of the relevant metrics embedded in the Web Calculator according to the implemented practices. However, as shown in Paragraph 1.1, this list is part of a considerable knowledge base we built, that allows us to easily extend the Web-based Calculator with additional practices.

Electricity (kWh)

Carbon emissions (g/kWh)

Capital Expenditures - CAPEX (€) o Hardware o Software o Service

Operational Expenditures - OPEX (€) o Electricity cost

Equipment Lifespan (years)

E-waste (kg)

Carbon emissions is used to estimate savings in CO2 emissions when adopting a green ICT practice. To obtain this value, the Electricity metric is multiplied by a number which represents the CO2 grams per KWh emitted in the electricity plants of the Netherlands (normalized average), namely 597 grams per KWh. This value has been calculated by means of the figures shown in Table 1.

VU University Amsterdam 2

Type of fuel Emissions (g/kWh) Plant Power (MWe1)

Gas 430 5400

Coal 900 3943

Nuclear 6 485

Renewable 0 37

Table 1 - Emissions and types of power plants in the Netherlands2

CAPEX is the amount of money (in Euro) to be invested in capital expenditures before the adoption of a green ICT practice. It can be further divided into Hardware, Software and Service (maintenance, periodic licenses) costs. OPEX is the amount of money (in Euro) to be spent on operational expenditures during the adoption of a green ICT practice. This includes Electricity cost that is derived from the Electricity metric according to the relative energy tariff. Equipment lifespan is used to express how long IT devices can be functional, and thus, when they should be replaced. The replacement will involve a periodical payment. E-waste expresses the quantity of IT material that has to be disposed of. This metric has an inverse relationship with the Equipment Lifespan metric.

2. The Web-based Impact Calculator for green ICT practices

The Web-based Impact Calculator for green ICT practices (http://greenpractice.few.vu.nl/index.php/calculator/step_1) is an online web application that helps decision makers to calculate the cost benefits of green ICT practices. To ease this achievement, the Web Calculator has been partially integrated with our online library of green ICT practices (http://greenpractice.few.vu.nl/) thus allowing users to immediately calculate an economical estimation of applying a specific practice in their organizations.

2.1. Implementation details

The Web Calculator has been developed as a PHP Web Application. Along with the PHP programming language, CSS stylesheets and JavaScript are used to improve user interface in terms of usability and aesthetics. For some specific functions, the application makes use of the following external libraries:

PHPExcel3: used to parse the Excel models of the practices (see Section 3 for details);

FileThingie4: used to manage and organize the models;

PHPlot5: used to draw charts to visualize the estimated values over time.

All these libraries are open source and available on the Internet. Up to now, the Thin Client practice from the green ICT library has been implemented in the Web Calculator. As part of our future work, additional practices will be also modeled and integrated.

2.2. Usage instructions

The Web Calculator consists of two modules: model management and selection-processing of models. The first module offers features to find, upload and retrieve models. The second module is divided in four sub-steps on which users select previously loaded models, configure them and calculate investments and expenses according to the modeled practice. In Figure 1 is shown the execution flow of the application. After the execution, charts of the results are available to users, to provide visualization of the economic benefits.

1 MWe = Million Watts electricity 2 Source: Renewable Energy in the Netherlands 2010, Statistics Netherlands, 2010,

http://www.manicore.com/anglais/missions_a/carbon_inventory.html, http://nl.wikipedia.org/wiki/Lijst_van_elektriciteitscentrales_in_Nederland 3 http://phpexcel.codeplex.com/

4 http://www.solitude.dk/filethingie/

5 http://phplot.sourceforge.net/

VU University Amsterdam 3

Model management

Models are represented as Excel spreadsheets. The file manager of the Web Calculator helps users to organize the different models in a corresponding structure for easier manipulation. By organizing these files, it is easier to select the specific models to be evaluated. To implement this functionality, File Thingie was chosen due to its simplicity, usability and support of general file management operations. In the list below you can find all the available file management operations of this manager:

Create folders

Rename folders/files

Delete folders/files

Moves folders/files

Upload/search files

In Figure 2 we show a screenshot of the Model Management interface.

Figure 1 - Usage diagram of the Web Calculator

Figure 2 - Models management screenshot

VU University Amsterdam 4

Selection and processing of models

Before running the models it is necessary to select and configure them. To do so, user selects a folder and the files containing the models of interest (see Step 1 and 2 in Figure 3 and Figure 4).

After the model selection, user is able to customize their labels (see Step 3 in Figure 5) and configure the type of the value exchange in the model (see Step 4 in Figure 6).

Figure 3 - Step 1: Selecting a source folder containing the models

Figure 4 - Step 2: Selecting the e3value models for processing.

VU University Amsterdam 5

Figure 5 - Step 3: Set labels for the e3value models.

Figure 6 - Step 4: Identify investments and expenses among metrics.

VU University Amsterdam 6

Each model has value exchanges which represent an action between the actors of the model. For example, the electricity provider (actor) exchanges electricity for money with the company (actor). These value exchanges can be either an Investment or a monthly Expense depending on whether exchanges occurs once or every month. In the example before, electricity is a monthly expense because companies calculate electricity as a monthly cost. After defining these properties users can process the models and visualize the results through charts. The most important part of this application is the one where models are executed. Users set up the parameters that are necessary for the processing of the selected models. For the Thin-client model such parameters can be the number of clients/servers, the cost of the equipment, the electricity consumption and the cost of electricity.

Once these parameters are set up, the application returns the useful metrics as a result. To get these results, models are loaded in the application as Excel spreadsheets. As a consequence, the Web Calculator provides support for reading spreadsheets by using the PHPExcel parser that reads the models and return the metrics. This library parses the spreadsheet (model) and loads the user-defined parameters into the equations of the spreadsheet which in turn calculates the results. The results show two types of potential savings: one time savings and monthly savings. In Figure 7, these metrics are shown as the investment savings (one time savings) and monthly savings for electricity consumption (in KWh), CO2 emissions and monthly costs.

Generate Charts

Last but not least, users are able to have a visualization of the metrics and values presented in the Final Step. To do that, the Web Calculator uses the PHPlot library that draws the charts resulted of processing the previous Thin/Fat models , as illustrated in Figure 8. Charts display the calculated metrics for a given period of time which by default is 12 months. A direct contrast between the two models, thin and fat client, show the difference which concerns monthly costs and payback time. Moreover, two additional visualizations demonstrate the electricity savings in KWh and the CO2 emissions savings. The period of time can be modified by the user and the charts are automatically updated.

Figure 7 - Final Step: Processed models.

VU University Amsterdam 7

3. Green ICT practice modeling

In this section we will describe how the economic models of the green ICT practices have been obtained. The modeling technique we have adopted is further described in a previous work [4] along with calculation examples.

3.1. The e3value methodology

The Web Calculator parses the models from Excel spreadsheets obtained through the e3value modeling tool. The e3value methodology, proposed by Jaap Gordijn [5], models enterprises and end-users exchanging objects of economic value, such as goods, services, and money, in return for other objects of economic value. In the following, we introduce the main concepts or constructs supported by the e3value modeling tool and their associated notations.

Actor. An economically, and often legally, independent entity. Examples of an actor include a customer, an organization and a company. In the notation, an actor is represented by a plain rectangle.

Market segment. A set of actors that share a set of properties. Actors in a market segment assign economic value to value objects equally. In the notation, a market segment is represented by a set of stacked rectangles.

Value port. Something that is used by an actor/market segment to provide or request a value object. In the notation, a value port is shown as a small arrow inside a value interface.

Value interface. Something that group value ports together and show economic reciprocity. Economic reciprocity means that actors/market segment will only offer value objects if they will receive value objects in return. In the notation, the value interfaces are drawn at the sides of actor/market segments as a thin rectangle with rounded corners, with value interfaces within.

Value exchange. Connect two value interfaces and represent a potential trade of value objects. In the notation, value exchanges are drawn as lines connecting the port of actors/market segment to each other.

Value object. Something that actors exchange which is of economic value for at least one actor. A value object is a service, a good, money, or an experience. Examples of value objects are products, delivery service and tuition fee. In the notation, a value object is represented as a label on a value exchange.

Figure 8 - Visualization of results.

VU University Amsterdam 8

Dependency path. The path where value exchanges, which is used to count the number of exchanges. In the notation, dependency path starts with a start stimulus and ends with a stop stimulus.

3.2. Modeling green ICT practices using e3value

Through e3value, we modeled all the practices implemented in the calculator. For this purpose, we needed the metrics elicited in the previous research activities: e3value constructs can be associated with numbers or parameters. We associated these parameters to the set of green ICT metrics needed for the economic estimation. For example, the desktop virtualization practice, one of the practices we implemented in the Web Calculator, is described in our library as

6:

A desktop virtualization software facilitates the use of thin clients7. These thin clients are far more energy efficient than regular fat

client computers. There is however an increase in server side computing due to the extra load of providing the desktops, which leads to an increase in energy consumption of servers.

Using the e3value modeling tool, we modeled an AS-IS situation (i.e. usage of fat-client without virtualization) and TO-BE situation (i.e. usage of thin clients with virtualization). Figure 9 shows the AS-IS situation, where company X purchases a number of fat-clients and servers from hardware suppliers in order to meet its computation needs, pays money to electricity suppliers for the electricity consumed by these fat clients and servers, and hosts an IT department (within the company or outsourced) to maintain the hardware devices ensuring they perform as expected. illustrates the TO-BE situation, where company X purchase thin clients rather than fat clients and the IT department has an additional task of providing and maintaining a virtualization software to deliver desktop virtualization service. From the models, the e3value tool generates net value flow sheets, which show the value exchanges flowing in and out of each actor. Net value flow sheets can be expressed in machine-readable formats (Excel sheets, RDF) or in human-readable graphs. The Web Calculator parses these spreadsheets and reconstructs the relationship between actors, in terms of value exchanges and metrics. After the user inputs the needed parameters for the metrics, the Calculator is able to perform the economic estimation of the benefit of the practice. In Appendix A, we provide all the e3value models relative to the practices we implemented.

6 http://greenpractice.few.vu.nl/index.php/action/view?id=25

7 i.e. workstations with minimal hardware configurations.

Figure 9 - Usage of fat clients, without virtualization.

VU University Amsterdam 9

4. Conclusions and future outcomes

The Green ICT Web Calculator provides an estimation of the economic impact of green ICT practices. The usefulness of this contribution is two-fold: first of all, it triggers a win-win mechanism for practitioners and researchers, providing a rationale for promoting sustainability in organizations as well as a stimulus to identify and formalize new green practices for achieving more profitable results. Secondly, it provides an educational tool to explain environmental benefits of re-greening ICT, as well as the relation with economic investments, gains, and ROI. The Web Calculator is easily extendible to include more practices, factors and metrics. As a matter of fact, we are already planning to integrate this tool with other similar tools independently developed by fellow researchers, aimed at modeling different aspects of green ICT (e.g., green networking). This also constitutes a great opportunity for interdisciplinary collaboration with our colleagues at the Faculty of Economics and Business Administration. Further research, of course, has to be carried out to validate the results generated by the calculator. We are already planning empirical experiments aimed at validating the energy savings gained by applying software-related green practices. These experiments will be carried out in collaboration with SURFnet and Green IT Amsterdam. In the future, the energy/environmental impact of companies will deeply affect their way of making business. This tool allows companies to increase their commitment towards green ICT, and this will create a positive feedback for more innovative green hardware and green technologies, and eventually for a sustainable society.

Bibliography

[1] Qing Gu, Patricia Lago, and Paolo Bozzelli, "A systematic literature review of green software metrics," ACM Comput. Surv., p. 31, 2013.

[2] Sven Gude and Patricia Lago, "A Survey of Green IT - Metrics to Express Greenness in the IT Industry," VU University Amsterdam, Tech. rep. aug 2010.

[3] Qing Gu, Patricia Lago, and Simone Potenza, "Delegating Data Management to the Cloud: a Case Study in a Telecommunication Company," in International Symposium on the Maintenance and Evolution of Service-Oriented and Cloud-Based Systems (MESOCA), vol. 7, sep 2013, pp. 56-63.

[4] Qing Gu and Patricia Lago, "Estimating the Economic Value of Reusable Green ICT Practices," in Safe and Secure Software Reuse.: Springer, 2013, pp. 315-325.

[5] Jaap Gordijn, Hans Akkermans, and J Van Vliet, "Designing and evaluating e-business models," IEEE intelligent Systems, vol. 16, no. 4, pp. 11-17, 2001.

Figure 10 - Usage of thin clients, with virtualization.

VU University Amsterdam 10

Appendix A. e3value models of green ICT practices

A.1. Power management

A.2. Energy-efficient monitors

Figure 12 - e3value model of Energy-Efficient Monitors practice

Figure 11 - e3value model of Power Management practice

VU University Amsterdam 11

A.3. Thin Client

Figure 13 - e3value model of Thin Client practice

A.4. Desktop Virtualization

Figure 14 - e3value model of Desktop Virtualization practice

VU University Amsterdam 12

A.5. Thin Client with PoE

Figure 15 - e3value model of Thin Client with PoE practice