A STUDY OF THE EFFECTIVENESS OF “WAKE UP ON LAN”

AS A MEANS OF POWER MANAGEMENT

Colin Pattinson and Linton Robinson

Innovation North Faculty of Information and Technology, Leeds Metropolitan University, Leeds U.K.

Keywords: Power management, wake on LAN, Energy savings in corporate IT.

Abstract: The growing awareness of the fragile nature of our environment, and of the damage that humankind is

causing to that environment, makes it necessary to review the environmental impact of all aspects of human

activity. One such area is the operation of corporate information technology (IT) systems. The growth in the

number and complexity of such systems over recent decades has led to a consequent increase in their power

requirements, to the point where for organisations of any size, “corporate IT” will be a major contributor to

the organisation’s overall energy consumption. A drive for overall reductions in consumption will, in the

words of the chair of the newly formed UK Environmental IT Leadership Team, mean “the IT department

becomes the focus of carbon reduction policies”. It is therefore timely to consider the present situation in

respect of power consumption within corporate IT systems, and to explore the potential avenues for

reducing that consumption. In this paper, we will briefly overview the area of power usage in IT, before

reporting on the outcomes of a specific project in which we explored the operation of the “wake on LAN”

method in a real situation.

1 INTRODUCTION

The rapid growth in computer use is placing a strain

on the UK’s ageing electrical infrastructure. The

growth of networked systems within organisations is

a major factor, with an increase in the number of

servers in a typical organisation from around 4 in the

late 1990s to 10 to 15 times that number (Watson,

2006). In October 2006, there were 2,507,278

registered companies in the UK (Companies House,

2006). While we do not know exactly how many of

these companies operated with the 60 servers in

Watson’s analysis or even as many computers, it is

not unreasonable to suggest that a good proportion

will have PCs, Servers and other network devices.

Data for energy consumption of IT systems vary

according to the exact definition of what is included,

and there is variation in reporting methods.

Kawamoto et. al. (2000) found “devices connected

to the Internet” for “commercial use” were

responsible for 2% of the total U.S. power

consumption. In Germany over 7% of domestic

energy consumption is “computer-related”

(Loerincik 2006).

One attempt to save energy has been to limit the

power taken by devices when not in use by placing

them in a “dormant” state, like the “stand by” mode

of many consumer electronics devices. Note: in a

further parallel with consumer electronics,

“dormant” is not “off”, and some power is still

required. The generic term for this activity is power

management (PM). PM can bring about energy

savings “…of the 74 TWh/yr of electricity

(approximately $6 billion per year) … consumed by

the Internet in the USA alone … 32% could be

saved with effective PM on desktops” (Christensen

et. al. 2007), and the equivalent of 250M litres of

gasoline per day could be saved if 1M PC users

employed PM (Long 2006).

Many current energy reduction activities

(including the one reported here) aim at the desktop

(user PC) element of IT systems, reasonably enough

in view of the numbers of such devices. However,

server and network device (switches, hubs and

routers) must not be overlooked. Allowing such

devices to enter a dormant state is problematic, as

typical network performance and security

monitoring techniques expect networked devices to

respond to a regular sequence of polling packets;

indeed some network management tools generate an

alarm if a device does not react to a probe request;

therefore there is a contradiction. Furthermore,

Pattinson C. and Robinson L. (2008).

A STUDY OF THE EFFECTIVENESS OF “WAKE UP ON LAN” AS A MEANS OF POWER MANAGEMENT.

In Proceedings of the International Conference on e-Business, pages 73-76

DOI: 10.5220/0001904300730076

 SciTePress

“[t]he immense increase in networking … is an

important factor [limiting the use of power saving]

because network activity can keep computers from

entering low-power modes” Webber et. al. (2001).

2 APPROACHES TO PM

In the opening parts of this paper, we argued that

there is a contradiction between the conventional

way networks operate, PM and cost savings. Let us

therefore take an overview of the different potential

strategies in this regard. We start by giving a brief

explanation of each of these techniques (Table 1).

To some extent, this table is rather simplistic;

e.g. overlaps between some techniques could

combine to effectively provide a PM solution. In

this study, WOL has been chosen primarily because

of its ease of use and simplicity in implementation.

Also, a previous research project successfully used

WOL in both wired and wireless scenarios, so we

have confidence in the underlying method.

3 QUESTIONS TO ADDRESS

The objective of this research was to determine how

WOL operates; hence we begin by exploring how

WOL will be delivered in a typical networked

scenario. Crucial to the deployment of WOL across

a network are questions of the nature and operation

of the communication process; for instance, are

WOL packets broadcast, unicast or multicast? What

are the issues with regard to a ‘connectionless or

connection-oriented approach?’ and finally, what are

the issues in terms of efficiency and security.

4 THE MAGIC PACKET (MP)

‘Magic Packet’ (MP) and ‘Wake up On LAN’

(WOL), are the same thing. The Magic Packet is

broadcast via port 7 or 9 which can be sent using a

variety of connectionless protocols, with UDP the

most common. (CapaInstaller, 2006). The Magic

Packet structure is similar to an Ethernet packet, but

the destination Medium Access Control (MAC)

address is repeated 16 times within the packet.

Table 1: Alternative PM approaches.

Dynamic Power

Management

(DPM)

Lu et. al. (2000)

Algorithms designed to shut

down a device only when an idle

period is of sufficient length to

justify performance degradation

and state-transition energy.

EZ Group Policy

Objects (GPO)

A free tool which allows network

administrators to centrally

control Power Management

settings using GPOs.

SMS Wakeup

(Hobbs 2006),

SMS 2003 currently lacks the

means for waking up machines, a

plug-in from 1E, Inc., called

SMS Wakeup, provides this

functionality. SMS Wakeup

integrates with SMS and uses

data from the SMS asset

collection to obtain MAC

addresses for each client

computer.

Using a low-power

channel

Shih et. al. (2002),

By adding a second low-power

channel, it is possible to shut the

system off and reduce idle

power. Out-of-band control

information can be sent

simultaneously, to maintain

connectivity and wakeup the

universal communicator (UCoM)

device when necessary.

Wake on LAN

(WOL)

(Korn et. al.,

2006).

Wake on LAN is a Layer 2-based

means for waking up machines

from sleep states such as system

standby, hibernate and shutdown

– and for remote access to them.

WOL encapsulates such Magic Packets inside a

broadcast UDP packet. This gives benefits over raw

Ethernet frames including Operating System (OS)

interoperability.

The complete UDP packet, sent over an Ethernet

interface, looks like this: (Spurgeon 2006)

[Ethernet header][IP header]

[UDP header][Magic sequence][CRCS]

5 TO BROADCAST, UNICAST OR

MULTICAST?

Unicast packets are sent from host to host, a one to

one relationship. Broadcast packets are one to many,

one host communicating with all other hosts.

Multicast is a restricted one to many relationship; a

ICE-B 2008 - International Conference on e-Business

Figure 1: Routed Network – MP Implementation.

single device can communicate with a specific set of

hosts. Our previous work with the MP showed that

it can be transferred by broadcast, unicast and

multicast, however for the purposes of this paper, we

consider a ‘broadcast’ approach, minimising effort

(just one message needs to be created), while

maximising coverage. By design, most modern

routers will block IP broadcast traffic and restrict it

to the local subnet, but this can be changed by the

network administrator.

6 WOL QUESTIONS

6.1 Efficiency

We conducted a series of tests to send the MP over a

wired network with unicast and multicast

instructions to wake machines. In these tests, the

packet successfully woke up and shut down

workstation(s). When this was carried out using a

wireless link, we found that in 2 cases, the

instruction to sleep succeeded, but the instruction to

wake failed if the target machine was in another

subnet. We believe that a correctly configured

wireless AP in that foreign subnet will rectify this

and conclude from these results that the MP is able

to produce satisfactory results and to do so reliably.

6.2 Reliability

It is well known that UDP is not reliable (Bhatti

2007); therefore the immediate question might be

why use it? Why not invoke a connection oriented

protocol with reliable delivery? Once a command is

issued to sleep, there is currently no feedback

mechanism to indicate that command was received

and obeyed. Instead, the network administrator may

need to utilise other tools creating extra work and

traffic, reducing the potential benefits of PM.

Therefore one challenge for developers is to bridge

the gap between the command and the response.

6.3 Security

There are a number of security issues within WOL

and its implementation (Robinson, 2007). For

instance, applying the ‘layered approach’ of most

network designs as in the following model:

((PM + current security) + (LAN security) +

(WAN security) + internet security))

= Total Vulnerabilities

means that vulnerabilities are compounded. We also

found that “more advanced systems” do not

necessarily mean “more advanced PM” a further

indication of the need for research and development

into PM. Min and Chandrasakan (2003) argue that

computational algorithms and low power digital

hardware trade energy for quality, because digital

processing occurs more slowly and uses less energy

when circuit voltage is reduced.

7 BROADCASTING THE MP

OVER ROUTED NETWORK

In section 5, we identified different ways of sending

the MP. Most networks will be routed in some way.

So we first consider if we can broadcast the MP.

Figure 1 shows a routed network with two broadcast

domains (10.10.40.0 /24 and 172.16.0.0 /16),

randomly selected and intentionally kept distant

from each other, (Class A and Class B respectively)

for ease in configuring the router.

It was possible to send the MP from A to B and

C, to shutdown and wake-up. We already knew that

A STUDY OF THE EFFECTIVENESS OF “WAKE UP ON LAN” AS A MEANS OF POWER MANAGEMENT

A to B was possible, from results obtained in the

previous test; however, here we successfully showed

that the MP does route to another domain. Though a

single router is used, in a network comprising of

many routers, the MP should also be able to achieve

the ends of Power Management, provided that

routers and firewalls are configured to allow

broadcast packets to travel between subnets (Korn

et. al., 2006). (The tool used in these experiments is

Prof Shutdown 3.2 evaluation version

(profshutdown, 2006)).

8 CONCLUSIONS

This paper shows that PM can be supported by MP

without detriment to network performance. Our own

research indicates that the adoption of PM take-up is

at best ad-hoc, sparse and often a small project after

business critical objectives have been realised. We

believe this to be counterproductive, as buy-in by all

stakeholders will itself contribute to the bottom line

of businesses and our world. Because the MP piggy-

backs Ethernet, we further believe that the case for

its deployment and the potential return on

investment (ROI) is significant.

REFERENCES

CapaInstaller (2006) www.capainstaller.com/index.asp

?page=157&page2=212&page3=269

Christensen, K., George, A.D. Nordman, B. (2007) The

Energy Efficient Internet Project. University of South

Florida. www.csee.usf.edu/~christen/energy/main.html

Companies House (2006) Verbal conversation

Hobbs C. (2006) Does SMS 2003 Support Wake-on-LAN?

http://www.faqshop.com/sms2003/sms2k3swd.htm

Profshutdown (2006) www.profshutdown.com/

(Accessed: December 2006)

Kawamoto, K., Koomey, J. G., Nordman, B., Brown, R.

E., Piette, M.A. Meier, A.K. (2000) Electricity Used

by Office Equipment and Network Equipment in the

U.S. Lawrence Berkeley National Lab document

LBNL-44917 August 2000

Korn D., Huang R., Bolioli T., Walker M. (2006)

Computer Power Management for Enterprises - A

Practical Guide. Proceedings of the 2006 IEEE

International Symposium on Electronics and the

Environment, May 2006 .pp. 161- 166

Loerincik, Y (2006) Environmental impacts and benefits

of information and communication technology

infrastructure and services, using process and input-

output life cycle assessment. PhD Thesis Ecole

Polytechnique Federale de Lausanne. September 2006.

Long C. (2006) Energy cost of PCs on standby. British

Broadcasting Corporation (BBC). news.bbc.co.uk/

go/pr/fr/-/2/hi/programmes/click_online/4929594.stm

Lu Yung-Hsiang, Chung Eui-Young, Simuni’c Tajana,

Benini Luca (2000) Quantitative Comparison of

Power Management Algorithms. Proceedings of the

conference on Design, Automation and Test in

Europe, March 2000 pp. 20-26.

Min R. Chandrakasan A (2003), Top Five Myths about the

Energy Consumption of Wireless Communication

Mobile Computing and Communications Review Vol.

7 No. 1 pp. 65 - 67 January 2003

Shih E., Bahl, P., Sinclair, M.J.(2002) Wake on wireless:

An event driven energy saving strategy for battery

operated devices. Proceedings of the Eighth Annual

International Conference on Mobile Computing and

Networking (MOBICOM'02), September 2002.

Robinson L. (2007) An Investigation into Power

Management and Power Savings: Is this achievable?

MSc Mobile and Distributed Computer Networks

Dissertation, Leeds Metropolitan University July

2007.

Rtsecurity (2006) www.rtsecurity.com/kb/tech/

wakeonlan.php (Accessed: 23/11/07)

Spurgeon C. (2006) Wake on LAN. http://gsd.di.

uminho.pt/jpo/software/wakeonlan/mini-howto/wol-

mini-howto-2.html

Watson J. (2006) Power shortage hits IT

www.computing.co.uk/computing/news/2150433/pow

er-shortage-hits

Webber C. A., Roberson J. A., Brown R. E., Payne C. T.,

Nordman B., Koomey J. G., (2001) Field Surveys of

Office Equipment Operating Patterns. Energy Analysis

Program. Lawrence Berkeley National Laboratory.

http://enduse.lbl.gov/info/LBNL-46930.pdf.

ICE-B 2008 - International Conference on e-Business