Projects

Project Agile Development

In this course for HvA informatics students a energy efficient cloud video streaming service is developed.

Call for cloud service providers to join our new research proposal

New project proposal: Applied research on Energy Efficient IaaS

The HvA is preparing a project proposal to generate knowledge on energy efficient IaaS services, with emphasis on the differences between IaaS distributions like OpenStack and CloudStack and consequences of software/architecture choices on energy consumption. The research hopes to contribute to knowledge about energy efficient IaaS services, cost-effective management of IaaS services and greening of the IT sector. In various discussions with industry experts IaaS providers were mentioned as a relevant target group given that:

  • IaaS providers pay for the energy bills of the hardware they are using and therefore have a direct interest in more efficient IaaS architecture;
  • IaaS providers use open source solutions such as OpenStack and CloudStack with a large online community, which can be adjusted and tested on energy efficiency in the SEFLab;
  • Above mentioned solutions supports various architectures and configurations for IaaS services: there is insufficient knowledge of the consequences of these architecture choices and other configurations on energy consumption.
  • Considering the rapid growth of IaaS services NOW is the moment to make energy efficient choices in order to avoid high energy costs in the future for IaaS providers.
  • Although most IaaS distributions are mature in terms of IT infrastructure, power management functionalities are still underdeveloped. Within the communities around these distributions there is a need for research and expertise in this area.

Call: Interested parties before December 2013!

At his moment the HvA is looking for companies (especially SMEs; preferably Dutch or with a presence in The Netherlands) interested in participating in the aforementioned research topic. This is possible in several ways ranging from active participation in formulating research questions, managing a subproject, and supervising student projects, to participating in workshops and events of the project. A brainstorm will be organized by early December, all potential partners and interested parties are invited to join this session.

For more information: Robert van den Hoed (r.van.den.hoed@hva.nl ), Bo Merkus (b.merkus@ hva.nl), Cleantech Research Programme, Amsterdam University of Applied Sciences. The full project proposal can be found here: 

Download (PDF, 68KB)

Power consumption comparison of system-level vs OS-level virtulaization

In this research we want to quantify and compare the power consumption of the overhead of system-level virtualization versus OS-level virtualization. The research question is: Is there a measurable difference in energy efficiency and resources used when running high performance applications in VMs versus linux containers?

The idea of the experiments is as follows: Execute a number of resource-intensive programs on the bare OS, inside a linux container, and inside a VM, and measure the performance and power consumption in each case.

Results (intermediate)

As expected,  the performance and power consumption of the experiments is roughly the same whether executed on the base OS or inside a linux container. VMs are still to be researched.

Partner: University of Amsterdam

Period: juli 2013 – december 2013

Contact: Karel van der Veldt, karel.vd.veldt at uva dot nl

Validation of energy profilers

The research is aimed at validating whether so called ‘profilers’ (software tools that can measure the energy consumption of a system or application) are accurate in their readings. We compared energy consumption measurements from the SEFLab setup with the profiler estimations with the test server running specific operations.  This allowed us to validate whether profilers can indeed be used to accurately measure the energy consumption of software. The motive for this research was to have validated tooling available for measuring the energy consumption of software.

We expected to come up with a list of profilers accurate enough to give a reliable picture of the energy consumption of a software application. In addition, we expected to indicate the error margin of the profiler estimations.

Results

During this experiment we found that most profilers, even those available commercially, are not suitable yet for measuring the energy consumption of software. The problems started with installing the profilers. In some cases heavy configuration was required which in the end did not provide the desired results, and in other cases it became apparent that the profiler was dependent on extra hardware to perform its tasks. We did find that Joulemeter is accurate for the Windows OS.  Though also requiring external hardware (a WattsUp PRO device), Joulemeter readings were fairly accurate compared to the SEFLab readings. For an example, see the figure.

SEFLab measurements and Joulemeter estimations of CPU power over time

One limitation of this research is that only the CPU has been taken into account. Though we acknowledge that other components play a role in the energy consumption, we did not manage to perform our experiments without influencing these other components. Given that our measurements would be distorted, we decided to exclude the other components from this research.

Partners: Centric, SIG

Period: February 2013 – September 2013

Contact: Erik Jagroep, e.jagroep  at centric dot eu

MRA Cluster Green Software

MRA Cluster Green Software is a R&D project on green software. Within this project research and validation will take place, in collaboration with the knowledge institutes and industry partners within the project team. The aim is to group knowledge and research results on the green software subject and force a breakthrough on innovation and sustainability. Within this project we will develop a new research infrastructure, tools and models and validate them in practice. This might results in new R&D projects, involvement of small and medium sized enterprises and possible spin-offs. Finally this project should contribute to the development of this specific green IT industry in the Amsterdam region.

Partners:

University of Amsterdam
– VU University Amsterdam
– SURFsara
– GreenIT Amsterdam
– SIG
– Equinix
– Schuberg Philis

Period: January 2013 – October 2014

Contact: Bo Merkus, b.merkus@hva.nl

More information: www.clustergreensoftware.nl

Funded by:

EU flag2colors     logo Kansen voor West kleur jpg

Power estimation of three real-life application scenarios for Bits-Nets-Energy research

In context of the Bits-Nets-Energy research performed by Green IT Amsterdam, UvA, SARA, TNO, SURFnet, and SURF funded by AgentschapNL, Green IT Amsterdam collaborated with the SEFLab to construct three real-life application scenarios and compare them on power consumption.

This allowed us to survey the energy usage characteristics of processors, memory, hard disks and other compo­nents when a server is asked to transfer data, to perform calculation-heavy operations, and to provide a remote application to an end user. The measurements were compared to the idle state of the server, and therefore show the additional power consumption resulting from a request for service.

In the ‘idle’ situation, the system uses 150.4 W of power. The breakdown over the components was as follows:

Server load

CPU 1

CPU 2

MB

MEM

HDD1

HDD2

FANS

150.4

 23.4

23.0

 26.6

 24.5

9.18

 2.69

 3.32

Table 1: Idle power usage (in W)

Scenario 1: data retrieval

The first scenario inspected was the transfer of data. A file was downloaded from the server using the FTP proto­col. The throughput was 7.8 Mbit/s, resulting in a download of a 1 GB file in 131 seconds. The server load increased by only 4% compared to the idle situation, mainly because CPU power consumption increased (11%). This is not significant, as the FTP server was running additional log­ging functionality.

Server load

CPU 1

CPU 2

MB

MEM

HDD1

HDD2

FANS

155.9

26.2

24.3

26.7

25.3

9.14

2.69

 3.32

Table 2: data transfer power usage (in W)

Scenario 2: interactive software

The second scenario consisted of running an application for the remote control of a desktop (Ultr@VNC), with a video file running on the server that was being viewed at a connected client computer. Playing this movie for the client increased the server’s power consumption by 35%. This was mainly the result of an increase in CPU power consumption by 91%.

Server load

CPU 1

CPU 2

MB

MEM

HDD1

HDD2

FANS

204.1

 47.7

42.9

 26.9

 26.7

8.83

 2.64

 3.26

Table 3: remote video playback power usage (in W)

Scenario 3: CPU-intensive processing

The third scenario was an illustration of CPU-intensive processing: the re-encoding of a video file in a different for­mat. By re-encoding an MP4 file to DivX, we simulated a CPU-intensive task. Re-encoding resulted in a 40% increase in power consumption as compared to the idle state, and a 110% increase in power consumption resulting from the CPUs.

Server load

CPU 1

CPU 2

MB

MEM

HDD1

HDD2

FANS

209.2

52.1

46.1

 26.8

 24.7

9.24

 2.68

 3.33

Table 4: processing re-encoding a video file (in W)


Partner
: Green IT Amsterdam

Period: April 2013 – May 2013

Contact: Bo Merkus, b.merkus@hva.nl

More info: http://www.surf.nl/nl/publicaties/Pages/TransportingBitsorTransportingEnergy.aspx

Result: Transporting Bits or Transporting Energy – does it matter – May2013.pdf

Green Software Practices

Empirical Validation and Assessment of Best Practices for Writing Energy-Efficient Software

The energy impact of software has been recognized as significant with respect to the overall energy consumption of its execution environment. Many researchers are currently working on sophisticated software power models, able to estimate and predict the energy consumption of software applications. However, this effort has not been translated yet into useful information for practitioners and developers.

The aim of this work is assessing the energy impact of a number of Green Software practices extracted from industrial practice. The assessment will be performed by means of an empirical experiment. The experimentation will be driven by the following research questions:

RQ1: what is the impact of each practice?

RQ2: what are the impacting factors in terms of energy consumption?

RQ3: what are the impacted resources in terms of energy consumption?

For the purpose of this experimentation, we will develop a number of simple software demonstrators, that will serve as test cases for each Green Software practice. These applications will be then executed in the SEFLab. During the experiment, we will gather two types of data: power consumption and resource usage data of the different hardware components. Then, we will perform statistical analysis on the data to answer our research questions. We will analyze, for each test case implementation, the software structures involved in correlation with the most used resources. From this comparison, we aim at extracting meaningful information to define an holistic, semantic view of the complex relationship between software and energy.

Partner: VU University Amsterdam

Period: September 2013 – March 2014

Contact: Giuseppe Procaccianti, g.procaccianti  at vu dot nl

SURFnet innovation grant 2012

The Software Energy Footprint Lab (SEFLab): Towards the application of green software

Summary

The SEFlab was awarded the SURFnet innovation grant 2012 to: 1) improve the measurement setup; 2) test the energy efficiency of the first software applications and 3) perform a stakeholder analyses on green software.

  1. To improve the measurement setup a team of graduation students and researchers from SIG and CleanTech worked together to develop a setup that allowed for high frequency power measurements of separate server components like the CPUs, hard drives, memory, fans, and main board. In addition various error analyses were performed and a lab manual was written.
  2. The first software applications tested in the SEFLab were the 5 most popular internet browsers: Internet Explorer, Chrome, Firefox, Safari and Opera. To this end the energy consumption of 4 scenarios were tested against the energy consumption of the system in idle. The results of the test (energy consumption in Joule) are summarized the table below. No clear conclusion could be drawn from these results because the energy consumption of the browsers differed over the scenarios.
  3. An internal and external stakeholder analysis was performed in which stakeholder from data centers, software developers and universities were interviewed. This led to the following conclusions:
    • Energy efficiency of software is relatively unknown but the energy savings potential is large: 30 to 90%.
    • There are no effective incentives to help industry reach this savings potential.
    • A common language is lacking between software purchaser, developers and manager to communicate about the efficiency of software.
    • At the three universities of Amsterdam (HvA, VU, UvA) there is interest to include the topic of green software in curricula.

Period: May 2012 – December 2012

Partners: SURFnet, SIG

Results:
Final report
SEFLab poster
Presentation SURF Symposium

SEFLab internship Marco van Veen

PasfotoIntroduction
I am Marco van Veen, and at the time of the assignment i was studying Electrical Engineering at the Hogeschool van Amsterdam. At the time when looking for a graduation internship this assignment caught my attention since i was both interested in computers in general and the challenge of developing a laboratory that measures the energy consumption at a component level of a server sounded like a very fun challenge. Together with Vincent Tseng i was chosen to realize this laboratory to allow the Software Improvement Group to do research on the energy consumption of the server. From february 2012 til august 2012.

Research assignment
The main research question was: “Is it possible to accurately measure the energy consumption of a server in such a way that differences in consumption of the software that runs on the server can be detected?”

Methods
Since the main research question was too broad  to start developing right away the question was split in various sub-questions which first had to be answered. The first thing was reading up on the subject and searching for prior research on power measurement methods, computer power management, and benchmark methods and software. Also the servers donated to us by the Software Improvement Group and Schuberg Philis were analyzed on a component level to find locations where the energy consumption of the most important components such as the CPU, the memory, hard disks and fans could be measured. With the information gathered from this research the various research questions were answered and the laboratory could be developed.

Results
 The main result of the assignment was a first version of the SEFLab, sensors were placed in the server and connected to circuitboard for filtering out noise and unwanted signals, this was connected to a data acquisition device that gathered data to analyze in real time on a separate computer. The laboratory was also tested and proved to be accurate enough to even see the power consumption of events such as when a key on the keyboard is pressed or the power consumption caused by starting up various pieces of software such as the OpenOffice Suite and Firefox.

Download the report for more information:
SEFLab internship report: Marco van Veen