ISO/IEC TR 30132-1:2016

TECHNICAL ISO/IEC TR
REPORT 30132-1
First edition
2016-09-15
Information technology —
Information technology sustainability
— Energy efficient computing
models —
Part 1:
Guidelines for energy effectiveness
evaluation
Technologies de l’information — Disponibilité des technologies de
l’information — Modèles informatisés à efficacité énergétique —
Partie 1: Lignes directrices pour l’évaluation de l’effectivité
énergétique
Reference number
©
ISO/IEC 2016
© ISO/IEC 2016, Published in Switzerland
All rights reserved. Unless otherwise specified, no part of this publication may be reproduced or utilized otherwise in any form
or by any means, electronic or mechanical, including photocopying, or posting on the internet or an intranet, without prior
written permission. Permission can be requested from either ISO at the address below or ISO’s member body in the country of
the requester.
ISO copyright office
Ch. de Blandonnet 8 • CP 401
CH-1214 Vernier, Geneva, Switzerland
Tel. +41 22 749 01 11
Fax +41 22 749 09 47
copyright@iso.org
www.iso.org
ii © ISO/IEC 2016 – All rights reserved

Contents Page
Foreword .iv
Introduction .v
1 Scope . 1
2 Normative references . 1
3 Terms and definitions . 1
4 Abbreviated terms . 1
5 Reference computing model for end-to-end energy effectiveness evaluation .2
5.1 Overview of computing models . 2
5.2 Reference computing model and energy effectiveness evaluation . 4
6 Holistic framework for evaluating the applicability of energy effectiveness
improving technologies . 6
6.1 Motivation . 6
6.2 Overview of holistic framework . 7
6.3 Considerations for evaluating the applicability . 7
6.4 Examples of energy effectiveness evaluation . 7
7 Guidelines for determining the energy effectiveness of a computing model .11
Annex A (informative) State of affairs for improving energy effectiveness of computing systems .12
Annex B (informative) Survey for calculating power consumption of the components in
computing models.19
Bibliography .21
© ISO/IEC 2016 – All rights reserved iii

Foreword
ISO (the International Organization for Standardization) and IEC (the International Electrotechnical
Commission) form the specialized system for worldwide standardization. National bodies that are
members of ISO or IEC participate in the development of International Standards through technical
committees established by the respective organization to deal with particular fields of technical
activity. ISO and IEC technical committees collaborate in fields of mutual interest. Other international
organizations, governmental and non-governmental, in liaison with ISO and IEC, also take part in the
work. In the field of information technology, ISO and IEC have established a joint technical committee,
ISO/IEC JTC 1.
The procedures used to develop this document and those intended for its further maintenance are
described in the ISO/IEC Directives, Part 1. In particular the different approval criteria needed for
the different types of document should be noted. This document was drafted in accordance with the
editorial rules of the ISO/IEC Directives, Part 2 (see www.iso.org/directives).
Attention is drawn to the possibility that some of the elements of this document may be the subject
of patent rights. ISO and IEC shall not be held responsible for identifying any or all such patent
rights. Details of any patent rights identified during the development of the document will be in the
Introduction and/or on the ISO list of patent declarations received (see www.iso.org/patents).
Any trade name used in this document is information given for the convenience of users and does not
constitute an endorsement.
For an explanation on the meaning of ISO specific terms and expressions related to conformity
assessment, as well as information about ISO’s adherence to the WTO principles in the Technical
Barriers to Trade (TBT) see the following URL: Foreword - Supplementary information
The committee responsible for this document is ISO/IEC JTC 1, Information technology, Subcommittee
SC 39, Sustainability for and by Information Technology.
A list of all parts of the ISO/IEC 30132 series can be found on the ISO website.
iv © ISO/IEC 2016 – All rights reserved

Introduction
The world is experiencing explosive growth of data from mobile client devices, cloud services, social
networks, online television, the Internet of things, big data and from traditional enterprise computing.
The growth of data has been accompanied by a growth in the energy usage and carbon footprint of
IT along with increased costs. Much research has been performed regarding energy management for
the last two decades, most focusing on the evaluating and improving energy efficiency of individual
components or systems such as processors, memory, wireless networks base stations, laptops,
supercomputers, data centres, handheld devices and so on. However, several disparate systems,
or systems of systems, collectively use energy to accomplish a given task and satisfy service-level
expectations. Consider, for example, someone who takes a photo with a smartphone and posts it to
a social network for their friends to view. Taking and transmitting the photo consumes energy from
the smartphone while the data transfer, processing and storage consumes energy too. Likewise, when
friends view the photo, that activity will consume additional energy. To improve energy effectiveness, it
is necessary to consider the end-to-end energy use of a task or service involving multiple systems.
The ISO/IEC 30132 series provides guidelines for the end-to-end evaluation of energy effectiveness of
a reference computing model and suggestions for determining the energy effectiveness of a computing
model. This document comprises guidelines for energy effectiveness evaluation, including a reference
computing model that includes end-to-end data transfer, processing and storage.
© ISO/IEC 2016 – All rights reserved v

TECHNICAL REPORT ISO/IEC TR 30132-1:2016(E)
Information technology — Information technology
sustainability — Energy efficient computing models —
Part 1:
Guidelines for energy effectiveness evaluation
1 Scope
This document establishes guidelines for improving the energy effectiveness for computing models.
Specifically, this document provides
— a reference computing model for evaluating end-to-end energy effectiveness,
— a holistic framework for evaluating the applicability of energy effectiveness improving
technologies, and
— guidelines for evaluating energy effectiveness.
2 Normative references
There are no normative references in this document.
3 Terms and definitions
For the purposes of this document, the terms and definitions given in ISO/IEC 13273-1 and the
following apply.
ISO and IEC maintain terminological databases for use in standardization at the following addresses:
— IEC Electropedia: available at http://www.electropedia.org/
— ISO Online browsing platform: available at https://www.iso.org/obp/
3.1
energy effectiveness
end-to-end total amount of data transferred, processed and stored per unit energy of a computing model
4 Abbreviated terms
ARP address resolution protocol
BNG broadband network gateway
DHCP dynamic host configuration protocol
DSL digital subscriber loop
DSLAM DSL access multiplexer
FTTN fibre-to-the-node
FTTP fibres-to-the-premises
© ISO/IEC 2016 – All rights reserved 1

HSPA high speed packet access
ICMP internet control message protocol
ICT information and communication technology
ISP internet service provider
NIC network interface card
OLT optical line terminal
ONU optical network unit
PON passive optical network
PtP point-to-point
QoS quality of service
UMTS universal mobile telecommunications system
WDM wavelength-division multiplexing
5 Reference computing model for end-to-end energy effectiveness evaluation
5.1 Overview of computing models
This subclause provides a survey of trends for various computing models.
In the traditional client-server computing model, clients are connected to servers via networks such
as the Internet. In this paradigm, a server is a computer system that selectively shares its resources,
whereas a client is a computer that initiates requests to a server in order to use its resources.
However, the emergence of new computing models such as cloud computing and the Internet of things,
along with new devices such as mobile phones, tablets, wearables and Internet connected sensors
introduce new considerations for both computing models and determining energy effectiveness.
Additionally, data is now transported over traditional wired networks and over high- and low-speed
wireless networks. Some client devices only support wireless networks.
Energy effectiveness has traditionally been viewed on a per device-category basis, but now it is
important to look at energy effectiveness from an end-to-end perspective that includes all the devices,
sub-systems and software that delivers a given set of functionalities (also known as a service).
New paradigms for the creation and use of data also affect energy effectiveness end-to-end. For
example, many users have their data on multiple devices and synchronization between client devices
and servers has become common practice. This means the same data may reside on multiple devices.
On the other hand, some applications retrieve and display data on the client device only as needed. This
scenario increases loads on networks, servers and storage, which may increase energy consumption.
The shift to mobile client devices and the shift to cloud computing, along with increases in the total
number of connected devices have driven a dramatic increase in the number of data centres and faster,
higher capacity networks with a corresponding increase in energy use, while client devices continue
to improve their energy effectiveness. Customer expectations for highly available, responsive services
may cause servers, storage and networking equipment to stay at high power states longer, possibly
conflicting with power management schemes and energy effectiveness goals. New technologies such as
push notifications also increase data and energy use across systems.
2 © ISO/IEC 2016 – All rights reserved

Therefore, energy effectiveness assessments should identify all of the energy consuming components
in the computing model. The energy effectiveness of networks is calculated using manufacturers’
data on equipment energy consumption for a range of typical types of equipment in networks. The
manufacturers’ data can include the amount of energy consumption in various states of equipment
such as idle, active and fully utilized state. This approach enables an overall model of network power
consumption to be constructed and provides a platform for predicting the growth in power consumption
[23]
as the number of users and access rate per user increase .
Figure 1 shows a high level representation of the network model of the Internet. In Figure 1, the Internet
is segmented into three major components: access network, metro network and core network with data
centres. The model is an abstract representation of the Internet and, as such, does not include much of
the fine detail of the Internet’s true structure and topology. The model does account for the typical hop
[24]
count for packets that traverse the Internet .
The refinement to include a more realistic representation of the Internet’s topology is ongoing. The
access network connects individual users to their local exchanges. Some of the typical access network
technologies, such as digital subscriber loop (DSL) to deliver packets through fixed-line telephone
service, fibres-to-the-premises (FTTP) installations to provide shared passive optical network (PON)
or a point-to-point (PtP) Ethernet connection. In a PON, a single fibre from the network node feeds one
or more clusters of users by using a passive optical splitter. An optical line terminal (OLT) is located at
the local exchange to serve many access modems or optical network units (ONUs) located at each user.
ONUs communicate with the OLT in a time division multiplexing, with the OLT assigning time slots
to each ONU based on its relative demand. In a PtP access network, each ONU is directly connected
to the local exchange with a dedicated fibre to the exchange. In areas where the copper pairs are in
good condition, a fibre-to-the-node (FTTN) technology may be used. This technology uses a dedicated
fibre from the local exchange to a DSL access multiplexer (DSLAM) located in a street cabinet close to
a cluster of users. A high-speed copper pair technology, such as very-high-speed DSL, is used from the
cabinet to the users. In areas where copper and fibre are not available or feasible, wireless can provide
Internet access. Technologies for the wireless access include WiMAX, high speed packet access (HSPA),
and universal mobile telecommunications system (UMTS). For wireless access, a wireless modem,
located in the user, communicates with a local wireless base station which, in turn, is connected to the
[23]
central office .
The central offices in a city are connected to each other and to other cities via the metro/edge network.
This network also provides connection points for Internet service providers (ISPs). The metro and edge
network serves as the interface between the access and core networks. The metro and edge network
includes edge Ethernet switches, broadband network gateway (BNG) and provider edge routers. Edge
Ethernet switches concentrate traffic from a large number of access nodes uplink to two or more BNG
routers. The edge switch connects to two or more BNG routers to provide redundancy. The BNG routers
perform access rate control, authentication and security services, and connect to multiple provider
[23]
edge routers to increase reliability. The provider edge routers connect to the core of the network .
The core network comprises a small number of large routers in major population centres. These core
routers perform all the necessary routing and also serve as the gateway to neighbouring core nodes.
The core routers of any one network are often highly meshed, but only have few links to the networks
of other providers. High-capacity wavelength-division multiplexed (WDM) fibre links interconnect
[23]
these routers and connect to networks of other operators .
© ISO/IEC 2016 – All rights reserved 3

[23]
Figure 1 — High-level network structure for the Internet
5.2 Reference computing model and energy effectiveness evaluation
Since there are many components in modern computing models and each has unique energy
effectiveness characteristics, it is difficult to calculate the energy consumption of services. Therefore,
this document uses a simplified generic reference computing model consisting of client devices,
network equipment and data centre equipment as shown in Figure 2. It is assumed that data centre has
computing resources such as server and storage. This document considers the following assumptions
for the simple evaluation of energy effectiveness of computing models.
— This document evaluates energy effectiveness of computing models from the view point of end-to-
end data creation, processing, storage, consumption and sharing.
— The model also assumes a data-oriented scenario where data moves back and forth between client
devices and data centres.
This document considers following components for evaluating end-to-end energy effectiveness while
considering application’s characteristics:
— client devices;
— network equipment;
— data centre equipment.
For survey for calculating power consumption of the components in computing models, see Annex B.
4 © ISO/IEC 2016 – All rights reserved

Figure 2 — Simplified generic reference computing model for energy effectiveness evaluation
From the perspective of computing the overall energy effectiveness, this document takes a very high-
level macroscopic view. At this level, it is considered that the transport network consists of multiple
wired and wireless network equipment. Client devices can take many forms including wearable
technologies, sensors and controls. It is noted that this document considers network layer equipment
in estimating the energy effectiveness of networks because the detail information about networking
equipment below network layer is difficult to obtain due to security and confidentiality. Thus, internal
network equipment are not taken into account. In case of data centre, border routers that connect
data centre to the Internet are taken into account. The internal routers within a data centre are not
considered. However, if the internal network information of clients, transport networks and data centre
are available, those information may be utilized in order to estimate the energy effectiveness more
accurately. Therefore, the end-to-end energy effectiveness evaluation can be performed by calculating
the energy effectiveness of the following components:
— client devices;
— transport network: in Figure 2, it is assumed that transport network includes access, metro/edge
and core network;
— data centre equipment.
The individual architectures that make up the high level components in the model are much more
complicated and the corresponding details and complexities are not within the scope of this document.
The energy consumption (watt-hours) of the end-to-end path in the reference computing model can be
calculated as follows:
EE=+ EE+
∑∑∑
E2Eclient_device,ni et_equip,djkatacentre_equip,
i jj k
The energy consumption measurement of components in the reference computing model can be
performed as follows.
— Basic assumption: each component is equipped with monitoring function, which can monitor and
measure throughput level of target equipment. When the monitoring value reaches a designated
threshold value, the amount of consumed energy is measured.
— Client devices and data centre equipment: monitoring function can measure the amount of power
consumption at the offered load to the equipment. Calculation of the energy consumption on client
© ISO/IEC 2016 – All rights reserved 5

device or data centre equipment can be performed by using measurement time, T, and power
consumption upon offered load as follows:
T
EP= offered_load
()
∑
equipequip
— Network equipment: network throughput can be calculated as doing sum of traffic load entered into
each network equipment. For this, the monitoring function can collect information about network
throughput such as offered load to measure the power information of the equipment. When the
network throughput reaches at target threshold value for power measurement, the monitoring
function gathers the information upon traffic load on each node and calculates the energy
consumption at the target traffic load. The total amount of power consumption on the network
equipment can be calculated by using the information upon each traffic load. The consumed power
on total network equipment can be the sum of the consumed power on each node. The power
measurement can be done depending on different throughput states (e.g. 10 %, 20 %, etc.). When
the value of traffic load is known, any network equipment can calculate power consumption value
of each node according to traffic load. It is noted that if calculating power consumption of each
network equipment, a practical way to calculate the network energy consumption may be to take
each segment, measure the total energy consumed over T, divide that by the total number of bytes
of data that traversed that segment over T and then multiply by the number of bytes of data for the
service that traversed that network segment over T.
T
EP= traffic_load
()
∑
net_equipnet_equip
Since the energy effectiveness evaluation of reference computing model only deals with end-to-end
data scenario, the end-to-end energy effectiveness can be calculated as follows.
Comparing product metrics allow consumers, enterprises and carriers to add energy effectiveness to
purchase criteria. A straightforward way to estimate the energy effectiveness of a network or telecom
system is to normalize its energy consumption to the amount of transmitted data in the test.
Let B be the end-to-end total amount of data transferred, processed and stored during the
E2E
measurement time, T, in the reference computing model and η be the overall energy effectiveness
E2E
calculated. Therefore, the end-to-end network energy effectiveness is calculated by dividing the total
amount of data, i.e. B by the total energy consumption of the computing model.
E2E
B
E2E
η =
E2E
E
E2E
6 Holistic framework for evaluating the applicability of energy effectiveness
improving technologies
6.1 Motivation
As the energy effectiveness of computing systems that use a large amount of electricity becomes an
important issue to network providers and operators, there is an increasing demand for developing
standards for evaluating and assessing the effectiveness and applicability of various energy
effectiveness improving technologies. This clause defines effective and holistic framework for assessing
the energy effectiveness improving technologies.
The rapid increase of energy consumption of ICT products accelerates the development of energy
efficient technologies. However, since energy efficient technologies in different dimensions may cause
unexpected side effects in total energy effectiveness of a computing system, it becomes necessary to
develop effective methods for understanding the inter-relationship among multiple energy efficient
technologies and evaluating energy effectiveness based on holistic view. This subclause also provides
motivation for holistic framework.
6 © ISO/IEC 2016 – All rights reserved

For state of affairs for improving energy effectiveness of computing systems, see Annex A.
6.2 Overview of holistic framework
This subclause provides an overview of holistic framework. When evaluating energy effectiveness
of computing models, it is generally necessary to consider multiple energy effectiveness metrics
due to the heterogeneity of components comprising the computing models. For example, the energy
effectiveness of computing node may be expressed in terms of CPU speed over consumed energy,
whereas the effectiveness of network node may be expressed in terms of network bandwidth over
consumed energy. Holistic energy effectiveness evaluation means that the evaluation is performed
considering multiple energy effectiveness indices simultaneously. The effectiveness indices may be
homogeneous or heterogeneous among them. The holistic methods allow easy comparison of holistic
energy effectiveness among multiple computing models or temporal trend. Also, it is possible to
understand how the change of values in energy effectiveness metrics contributes to the overall energy
effectiveness. For example, if the operator of a data centre wants to increase entire data centre energy
effectiveness by double, it is possible to calculate how much improvement should be performed for each
metric. During the holistic evaluation, it is possible to consider the characteristics of each computing
model when estimating the holistic energy effectiveness using multiple energy effectiveness metrics. It
is noted that depending on the service level objectives and agreements between customers and service
providers, the performance and availability of services may vary. These service level objectives may
affect the energy consumption of customer services. Thus, it is desirable to take account of customer
expectations in order to provide more accurate energy effectiveness estimation.
6.3 Considerations for evaluating the applicability
This subclause provides considerations for developing and evaluating the applicability of energy
efficient technologies for computing systems during sustainable ICT product during use stage of
life cycle. In the computing models, energy is consumed during computing, data transport or data
storing operations. In other words, not only data transport, but also data manipulation causes energy
consumption. Moreover, the rapid deployment of mobile devices enables users to use multiple personal
devices and maintain synchronized data among the devices. Therefore, user’s activities can cause more
effects to the energy consumption of ICT infrastructure than legacy environments. Current computing
models impact energy requirements in multiple systems. Also, use of enterprise as computing, data
source and maintenance requires data management across multiple systems. Low power, low capability
end devices require increases in enterprise services and availability on resilient networks, servers and
storage, and low capability end devices increases data centre and network energy requirements.
Therefore, the evaluation of computing models needs to consider the following issues:
— consideration for application’s impact on ICT infrastructure in terms of data transport, computation
and data storing;
— consideration for application’s characteristics when executing the application. For example, users
may execute application in a local node that has relatively, but does not require data transport over
network or may execute the application at remote cloud node that has high performance, but needs
data transmission over networks.
6.4 Examples of energy effectiveness evaluation
As investigated in Clause 5, the energy effectiveness of end-to-end communication is represented as
a ratio of total amount of data transferred, processed and stored to total energy consumption. In this
subclause, it is described how to evaluate end-to-end energy effectiveness by using examples. Figure 3
shows an example scenario for end-to-end energy effectiveness evaluation. In this example, a client
device sends data traffic to the server in a data centre, and the sever stores the received data to an
external storage systems.
© ISO/IEC 2016 – All rights reserved 7

Figure 3 — Example scenario for end-to-end energy effectiveness evaluation
This document considers the following assumptions.
B = is the amount of data traffic (Mbytes) that client device transmits to the server in the data centre.
T = is the data traffic transmission time.
S = is the data traffic storing and retrieving time of storage.
∆S = is the time interval between storing and retrieving the data from storage.
N = is the number of hop counts in transport network.
P = is the average power consumption rate for client device during T.
AVG_client
P = is the average power consumption rate for router j during T.
AVG_router, j
P = is the average power consumption rate for server during T.
AVG_server
P = is the average power consumption rate for storage to store or retrieve data B during S
AVG_stor_acs
P = is the average power consumption rate for maintaining data in storage for ∆S.
AVG_stor_maintain
It is noted that in case that P , P , P , P and P are not
AVG_client AVG_router, j AVG_server AVG_stor_acs AVG_stor_maintain
available, P , P , P , P and P may be used as an
MAX_client MAX_router, j MAX_server MAX_stor_acs MAX_stor_maintain
approximation.
P =is the maximum power consumption rate for user equipment.
MAX_usr
P = is the maximum power consumption rate for router j.
MAX_router, j
P = is the maximum power consumption rate for server.
MAX_server
P = is the maximum power consumption rate for accessing storage.
MAX_stor_acs
P = is the maximum power consumption rate for maintaining data in storage.
MAX_stor_maintain
According to the description of end-to-end energy effectiveness presented in Clause 5, the maximum
energy effectiveness, η , can be expressed as follows. It is noted that this document refers only to
E2E
computing model dedicated to a single data flow. Also, server is assumed as a dedicated server to a
single application.
8 © ISO/IEC 2016 – All rights reserved

B
E2E
η =
E2E
E
E2E
B
=
N
EE++EE+
∑
client_devicerouter,j server storage
j=1
B
=
T 3 T T S
PP++ PP+ +
∑∑∑∑ ∑
AVG_client AVG_router,Aj VG_serverAVG__stor_acs
00j=1 0 0
ΔS
P
∑
AVG_stor_maintain
Figure 4 shows another example for e-mail exchange among users.
Similar to the previous example, the following assumptions are considered.
B = is the size of an e-mail message in Mbytes including attachments that user 1 sends to user 2 and
user 3. It is assumed that the users use an e-mail server located at the data centre.
T = is the data traffic transmission time.
S = is the data traffic storing or retrieving time of storage.
∆S = is the time interval between storing and retrieving the data from storage.
N = is the number of hop counts in transport network between the user 1 and the mail server in
usr1
data centre.
N = is the number of hop counts in transport network between the user 2 and the mail server in
usr2
data centre.
N = is the number of hop counts in transport network between the user 3 and the mail server in
usr3
data centre.
P = is the average power consumption rate for user equipment during T.
AVG_client
P = is the average power consumption rate for router j during T.
AVG_router, j
P = is the average power consumption rate for mail server during T.
AVG_mail_svr
P = is the average power consumption rate for storage to store or retrieve data B during S.
AVG_stor_acs
P = is the average power consumption rate for maintaining data in storage for ∆S.
AVG_stor_maintain
© ISO/IEC 2016 – All rights reserved 9

Figure 4 — Example scenario for e-mail exchange among users
According to Figure 4, it is possible to identify three data transport paths in this e-mail exchange
example.
① Deliver an e-mail message to mail server and store the e-mail to storage that are located at
data centre.
② Deliver the e-mail message from the mail server to user 2 equipment using routing path 2.
③ Deliver the e-mail message from the mail server to user 3 equipment using routing path 3.
Therefore, by summing up the energy consumption of the three data transport paths, it is possible to
approximately calculate the total energy consumption of the e-mail exchange scenario.
Similar to the previous example, the energy effectiveness of the example scenario can be calculated as
follows.
10 © ISO/IEC 2016 – All rights reserved

B
E2E
η =
E2E
E
E2E
B
=
EE++ EE+
∑∑
client_device,i router,sj erverstorage
i j
B
=
Energy Consumption of each routing path ()i +
∑
Energy CConsumption of storage equipment
∑
B
=
T 3 T T S
PP++ PP+
∑∑∑∑ ∑
AVG_usrA1 VG_router,j AVG_mail_svr AVG__stor_acs
00j=1 0 0
B
=
T 3 T T S
PP++ PP+ +
∑∑∑∑ ∑
AVG_client AVG_router,Aj VG_serverAVG__stor_acs
00j=1 0 0
ΔS
P
∑
AVG_stor_maintain
B
=
T 3 T T S
PP++ PP+ +
∑∑∑∑ ∑
AVG_client AVG_router,Aj VG_serverAVG__stor_acs
00j=1 0 0
ΔS
P
∑
AVG_stor_maintain
7 Guidelines for determining the energy effectiveness of a computing model
This clause provides the basic steps to evaluate energy effectiveness of a computing model in terms of
requirements, procedure, metrics and measurements. The general procedure for energy effectiveness
determination is as follows.
a) Identify the target computing model.
b) Identify the components of the target computing model. The components can be explained in terms
of computation, data transport and data storage.
c) Determine the energy effectiveness metrics for the components in the target computing model.
d) Determine the energy effectiveness of the target computing model from end-to-end perspective.
e) Determine the energy effectiveness of the target computing model using the holistic method
described in Clause 6.
© ISO/IEC 2016 – All rights reserved 11

Annex A
(informative)
State of affairs for improving energy effectiveness of computing
systems
A.1 Taxonomy
Energy-effectiveness certainly is an important issue in networks and computing systems. This annex
investigates various potential technologies for improving energy effectiveness of computing systems
so that it provides a taxonomy of energy consumption points for computing systems in the computing
models investigated in Clause 5.
As the necessity for considering energy consumption in computing and network systems increases, a lot
of research has begun to investigate energy awareness in networks and systems. This annex presents
taxonomy of energy consumption points for computing and network systems based on the computing
models investigated in Clause 7.
A.1.1 Classification criteria
It is generally known that the energy consumption point of computing models can be divided into
computing systems and networks. Thus, separate criteria for computing systems and networks are
described.
This document refers the criteria for classifying energy effectiveness of computing systems and
networks presented in Reference [2]. The general criteria for classifying energy effectiveness of
networks include timescale, scope of required information, networking layer, input process and
approach.
A first important criterion deals with the timescale of the decisions involved by the green strategy.
As described in Reference [2], timescales on the order of nanoseconds to microseconds can be applied
to CPU and instruction level, which is relevant in the computer and software architecture levels so
these timescales concern individual components of a single system. Timescales on the order of micro to
milliseconds deal with the system layer. At these timescales, actions may be taken between consecutive
packets of the same flow, possibly involving several components at the same time, but likely confined
within a single system. Larger timescales, on the order of one second and above, allow instead the
action to span between multiple systems, possibly involving coordination of such systems as well. In
order to present simple classification, it is possible to divide timescales into online and offline. Online
requires actions to be taken during the operation of systems, which takes typically less than several
seconds, whereas offline requires more time and the actions to be taken before runtime of systems as,
for instance, during the system design process.
The second criterion concerns the type and amount of components or systems involved. In other words,
it can be divided into local or global depending on the scope of the information required to take a
decision. Local strategies will require information that pertains to a single system, or single network
link, whereas global strategies will require information that pertains to a set of systems and links.
The third criterion is related with networking layer. Since most network equipment is implemented
with layering concept, it is possible to classify energy consumption points according to the networking
layer to which they apply. Based on the TCP/IP protocols stack, each energy consumption point can
either be classified into data link, network, transport and application layers, or may require cross-layer
interaction.
12 © ISO/IEC 2016 – All rights reserved

Another classification criterion comes from the analysis of the input process that drives the decision
taken in the solutions. The decision may be taken based on the instantaneous situation, on the historical
pattern or on the forecast. In the case of online solutions, all the three kinds of decision are possible. For
the offline solutions, historical pattern and forecast based analysis are only applicable.
Finally, another criterion related to the evaluation methods of the proposed technology is required.
Typical examples of the evaluation methods are discrete event simulation, hardware/software
prototyping or formal models with analytical or numerical solution. All methods have their respective
advantages and disadvantages. Theoretical analysis and simulation can be performed with low cost
and short time compared to prototype, hardware or real deployment. However, the latter approaches
may provide more precise evaluation and a higher level of maturity of a specific research field. Table A.1
[2]
shows the summary of the classification criteria .
Table A.1 — Summary of the classification criteria for energy efficient technology
Criterion Category Meaning
Defines the update frequency of
Timescale Online, offline
the policy.
Influences the volume of
Scope Local, global communication required to
reach the objective.
Link, network, transport, application, Individuate which entities
Networking layer
cross-layer shall collaborate.
Defines learning and adaptation
Input Instantaneous, historical, forecast
capabilities of the algorithm.
Traffic analysis, theoretical modelling, Flavour of the study, also
Evaluation approach simulation, hardware and software correlates with the level of
prototyping maturity of the work.
A.1.2 Taxonomy of energy efficient technologies
This subclause presents taxonomy of energy efficient technologies for computing models according
to the classification criteria investigated in A.1.1. As discussed, computing models can be divided into
computing systems and networks. Thus, the taxonomy consists of three categories, namely computing
systems, networks and storages, respectively.
Technologies for computing systems are categorized as follows.
— Hardware-level technologies
— Dynamic voltage scaling: a technology that allows the voltage supplied to some computer
components to be raised or lowered dynamically according to power, performance and thermal
requirements.
— Dynamic frequency scaling: a technology that allows the CPU clock speed to be automatically
adjusted according to power and thermal requirements.
— Link power management: technologies that allow parts of connected devices in a computer or
connected to a computer to enter low power states when idle.
— Energy efficient display: a computer display or monitor that uses energy saving technologies
such as organic light emitting diode (OLED) or techniques to reduce power consumption when
in sleep mode.
— Energy efficient power supply: a power supply which meets specified energy targets set by
organizations such as 80 PLUS.
© ISO/IEC 2016 – All rights reserved 13

— Energy efficient battery charger: a battery charger for cellular phones, laptop PCs or tablets
that incorporates energy saving techniques.
— Software-level technologies
— Workload consolidation: a technique that allows running multiple tasks on the same physical
machine in order to reduce the number of nodes that are switched on. A key component of
systems that aim to consolidate workloads is to monitor and estimate the load posed by user
applications or estimate the arrival of user requests.
— Energy-aware task scheduling: energy aware task schedulers have three types, namely offline
scheduling based on a prior task information, online scheduling which is purely dynamic and
hybrid approaches, including an offline phase where the slack is greedily absorbed and dynamic
algorithms operating in the online phase.
— Virtualization: creates a virtual machine that acts like a real computer with an operating system
and improves energy efficiency.
— Virtual machine migration: moves a running virtual machine or application between different
physical machines without disconnecting the client or application.
Technologies for networking systems are categorized as follows.
— Adaptive layer-2 technology: most current types of network equipment show a constant power
consumption profile largely irrespective of their actual utilization. The computing world on the
other hand has long embraced methods to approximate energy-proportional computing. In the
networking world, there are initial steps into energy-proportional communications. Adaptive link
rate technology is such a step where the energy use changes with link utilization.
— Network interface proxy technology: proxying describes technologies that maintain network
connectivity for other devices so that these can go into low power sleep modes. This mainly targets
the reduction of unnecessary energy waste through edge devices.
— Energy-aware infrastructure technology: in order to achieve a further reduction in energy use,
coordination and management of larger parts of the network appears to be a promising idea.
Energy-aware infrastructure describes a class of techniques to this end. Energy-aware routing is
one example that falls into this category. Energy-aware routing makes use of the fact that traffic
follows certain patterns. Based on this knowledge, in times where network traffic is low, a number of
routers can be put to sleep while the network as a whole still preserves connectivity and an adequate
service level. Requirements for an energy-aware control planes are outlined in Reference [5].
Tech
...