GENERAL RESEARCH

Technologies for the Electronic Distribution of InformationServices - A Value Proposition Analysis

Jochen Wulf & Ruediger Zarnekow

Received: 9 June 2009 /Accepted: 21 December 2009 /Published online: 2 February 2010# Institute of Information Management, University of St. Gallen 2010

Abstract Since the original design and deployment of theInternet architecture, the economical and technologicalrequirements regarding the distribution quality of web-basedinformation services have changed drastically. Businessmodels have evolved that particularly address quality andcost aspects of information service distribution, e.g. contentdelivery networks and peer-to-peer distribution. In addition,network operators apply differentiated routing technologies indedicated infrastructures to guarantee a superior quality ofservice (QoS). This article compares the value propositions oftechnologies for information service distribution. 103 infor-mation services were analyzed by means of discriminantanalyses in order to identify the main aspects influencingdelivery quality and costs. The results indicate that the valuepropositions differ with regard to the type of services theysupport rather than with regard to direct QoS criteria, such aslatency and packet loss. The insights derived from this worksupport information service vendors in their choice of adistribution provider.

Keywords Electronic distribution . Content delivery .

Quality of service

JEL classification L86—Informationand internet services .

Computer software

Introduction

The electronic distribution of information services has beensubject to a discourse in research since the immense growthof the Internet in the late-1990s. In e-business research, arich variety of issues concerning the application of web-based information services to support business processeshave been addressed (Amit and Zott 2001; Timmers 1998).

Physical distribution is defined as the collective term for therange of activities involved in the movement of goods from thepoint of production to the final point of sale, such aswarehousing and transportation (McKinnon 1988, 33). Theobjective of distribution is to meet customer desires withregard to the amount of delivered goods and the time and placeof delivery. Distribution providers must thereby optimize thebalance between quality and costs of delivery (Bowersox et al.1986, pg. 19). Even though the economics of informationservices are fundamentally different (Demirkan et al. 2008;Shapiro and Varian 1999), the objectives for the distributionof tangible goods also hold for the distribution of informationservices over communication networks (electronic distribu-tion). Despite the richness of e-business research, theeconomic aspects associated with technologies for electronicdistribution have scarcely been addressed.

Several authors (Faratin 2007; Leighton 2009; Saroiu etal. 2002) differentiate between three major distributionmethods in the Internet: centralized hosting, contentdelivery networks, and peer-to-peer file sharing systems.They describe technical implementations and find signifi-cant differences in traffic characteristics, but do notcompare attributes related to distribution performance. Inthis paper, a fourth distribution method, called directhoming, is taken into account which is commonly used inpractice and is based on proprietary IP networks or amodification of the public Internet architecture (Huston

Responsible editor: Axel Hochstein

J. Wulf (*) :R. ZarnekowFachgebiet Informations- und Kommunikationsmanagement,Technische Universität Berlin,Sekr. H 93, Strasse des 17. Juni 135,10623 Berlin, Germanye-mail: jochen.wulf@tu-berlin.de

R. Zarnekowe-mail: ruediger.zarnekow@tu-berlin.de

Electron Markets (2010) 20:3–19DOI 10.1007/s12525-010-0027-x

2000, pp 399–406, ISO 1989, Xiao and Ni 1999): in directhoming scenarios, network operators reserve dedicatedinfrastructure or perform traffic differentiation in order tomeet high performance requirements of specific services,such as IP TV.

Electronic distribution has evolved into a large commer-cial market. According to an industry report (FROST &SULLIVAN 2008), the content delivery network (CDN)market earned over $700 million in 2007, and is predicted toincrease earnings by up to 400% by 2013. A highly ratedtarget segment for CDN providers is the delivery of Video onDemand content, i.e. the streaming of multimedia contentfrom distributed servers. Other companies promise the samevalue proposition as CDN providers, i.e. high speed, reliabilityand efficiency, based on the peer-to-peer (P2P) distributiontechnology (Androutsellis-Theotokis and Spinellis 2004). P2Ptechnology is, for example, used to distribute IP TV contentin quasi real-time to global audiences.

From the perspective of an information service provider,the specific differences between the value propositions of thefour mentioned distribution methods remain largely unclear.The CDN provider Akamai and the P2P distributor BitTorrent,for example, both target download and streaming delivery(BitTorrent 2009; Akamai 2009). An exemplary service levelagreement of Akamai’s CDN service does not quantifyperformance improvement guarantees (Onecle 2009). Neitheracademic nor non-academic literature provides appropriatedecision support for the choice of a distribution method andstates which service requirements are met by whichdistribution method.

The objective of this paper is to compare the differenttechnologies for IP based distribution, and to outline theirdistinct characteristics and their abilities to provide value inspecific application contexts (value propositions). 103information services were analyzed in order to deduce thedistinctive value propositions of their associated distributionmethods. We argue that information services reveal inherentcharacteristics of the underlying distribution methods.

This article is structured as follows: in Chapter 2, therelationship between information service quality and ser-vice distribution is described. In Chapter 3, four distinct IPbased distribution methods, i.e. Centralized Hosting, Direct

Homing, Content Delivery Networks and P2P Distributionare explained. In Chapter 4, a set of information servicecharacteristics, which potentially bear a relationship todistribution performance, are identified. These relationshipsare verified by performing discriminate analyses in Chapter5. The article concludes with a summary in Chapter 6.

Information service quality and distribution

Service science researchers have identified a set of constitu-tive characteristics of services (Demirkan et al. 2008, amongothers). These characteristics lead to specific requirements onservice distribution and explain its importance (Fig. 1).

Firstly, the simultaneity of service production and con-sumption inhibits the preproduction and storage of servicesand imposes high quality requirements, e.g. concerning real-time service delivery. Secondly, the wide range of potentialservices results in heterogeneous requirements for servicedistribution. This makes it difficult to find a standard solutionfor the distribution of a service. Thirdly, the overall evaluationof the service quality as experienced by a consumer is, to alarge extent, subjective. For this reason, requirements for asatisfactory service delivery can not entirely be generalized. Inaddition, service delivery requires a direct customer involve-ment. As a consequence, the distribution quality, to a certaindegree, depends on a customer’s involvement (e.g. on theaccess capacity the customer allocates). Due to the intan-gibility of services the overall distribution quality can onlybe measured at the customer end, because this is thelocation at which a service is consumed. Intermediatemeasurements only allow predictions of overall servicequality. Lastly, services are perishable. That is why technicalparameters of distribution quality can only be measured atthe exact time of service provisioning.

The general concept of service quality has been subjectto extensive research in business sciences. It can beassessed by comparing a user’s expectations and percep-tions of the performance level for a range of serviceattributes (Parasuraman et al. 1985). Accordingly, theInternational Organization for Standardization states: Thequality of something can be determined by comparing a set

Service Characteristics

• Intangibility

• Perishability

• Customer contacts

• Simultaneity

• Heterogeneity

• Demand fluctuation over time

• Customization

• Complexity

Consequences for Service Distribution

• High requirements on service distribution quality (e.g. concerning real-time capabilities)

• Heterogeneity of distribution requirements and exchanged data

• Subjectivity of distribution quality experience

• Distribution quality depends on customer involvement

• Overall distribution quality can only be measured end-to-end

• Technical distribution quality parameters can only be measured during service provisioning

Fig. 1 Service characteristicsand consequences for servicedistribution

4 J. Wulf, R. Zarnekow

of inherent characteristics with a set of requirements [...] Aquality characteristic is tied to a requirement and is aninherent feature or property of a product, process, orsystem (ISO 2005). A requirement is understood as anexpectation or a need of a user. Various authors discuss thetypes of requirements which are to be taken into account inan information service quality assessment (Liao andCheung 2008; Liu and Arnett 2000; Zeithaml et al. 2000;Zeithaml et al. 2002). Zeithaml et al. (2002), for example,define the following attributes: information availability andcontent, usability, privacy/security, graphic style, andfulfilment. Such attributes represent the factors relevant toa user’s perception of service quality.

In telecommunications research, the term quality ofservice (QoS) is used heterogeneously to describevarious concepts of service quality. As discussed byGozdecki et al. (2003), the term QoS is used, on the onehand, to describe the customer’s service quality assess-ment (business oriented view) and, on the other hand, todefine technical parameters of service and networkperformance (technology oriented view). Figure 2 depictsboth views. Externally, i.e. in the relationship betweenservice provider and user, service quality describes thegeneral comparison of inherent service characteristics asexpected and perceived by users. Internally, i.e. among theparties involved in service production, service perfor-mance comprises all performance related parameters of aservice described in technical terms, such as speed,accuracy, availability and reliability. These parametersare affected by the inter-working of server, distributionand client systems.

The quality of service distribution represents onecomponent of service performance, as depicted in Fig. 2.It is determined by all systems involved in the delivery ofcontent from the originating servers to the end userterminals, such as routing, forwarding and caching systems(Tanenbaum 2003, pp. 343 et sqq.). Similar to the networkperformance concept (Xiao and Ni 1999; Zhao et al. 2000),it can be characterized by four quality parameters: band-width, delay, packet loss and jitter. Bandwidth defines theeffective volume of data per time unit being transmittedbetween communication end points. Delay comprises thelength of time that a data package takes from the sender tothe recipient. Packet loss defines the number of datapackets that are lost in the transmission from the sender tothe receiver. Jitter describes the fluctuation in the delay.Since it is a function of delay, it is not separately taken intoaccount in the following analyses. Several authors defineperformance levels for classes of services (CoS, Marchese2007, pp. 5–8, Gozdecki et al. 2003). CoS concepts definebounds for the QoS parameters to precisely describe what isconsidered an appropriate quality for specific serviceclasses, such as real-time conversational or near real-timeinteractive services.

Technologies for information service distribution

Different technological approaches have evolved since thedevelopment of the Internet architecture to supportinformation service distribution. These technologies aregenerally classified into four categories, summarized in

Fig. 2 Quality of service distribution and service quality

Technologies for the Electronic Distribution of Information Services 5

Tab

le1

Inform

ationservicedistribu

tiontechno

logies

andcase

exam

ples

Distribution

metho

dDefinition

Key

characteristics

References

Businessmod

elcase

exam

ples

Centralized

Internet

Hostin

gCentralized

internet

hostingrefers

tothe

useof

“one

orasm

allnu

mberof

collo

catio

nsitesto

hostcontent”

(Leigh

ton20

09).

-central

server,multip

leclients

-Leigh

ton(200

9)-bluehost(W

ebHostin

g)

-Internetbased

-Tanenbaum

(200

3,pg

.4)

-GoDaddy

(Web

Site

Hostin

g)

-Und

erwoo

d(200

1)-H

ostGator.com

(Dom

ainHostin

g)

-1&1(W

ebHostin

g)

DirectHom

ing

Directho

mingisthedistribu

tionof

services

vianetworks,which

are

able

toprov

idehigh

QoS

throug

htechno

logies

such

asresource

reservationor

trafficprioritization

(XiaoandNi19

99).

-proprietary

(mostly

sing

le)network

-Huston(200

0,pp

399–

406)

-AT&T(U

-Verse)

-con

figu

redaccess

lines

-XiaoandNi(199

9)-Swisscom

(Bluew

in)

-QoS

throug

hcapacity

reservation

orpacket

differentiatio

n-Zhaoet

al.(200

0)-BT(BTVision)

-T-H

ome(T-Entertain)

CDN

baseddistribu

tion

“CDNsactas

trustedov

erlaynetworks

that

offerhigh

-perform

ance

deliv

ery

ofcommon

Web

objects,static

data,

andrich

multim

edia

contentby

distribu

tingcontentload

amon

gserversthat

arecloseto

theclients.”

(VakaliandPallis

2003

)

-multip

leservers,multip

leclients

-Dilley

etal.(200

2)-A

kamai

(App

lication

Perform

ance

Solutions)

-Internetbased

-Pallis

andVakali(200

6)-A

mazon

(Cloud

fron

t)

-PathanandBuy

ya(200

8)-Lim

elight

(Lim

elightDELIV

ER)

-VakaliandPallis

(200

3)-Edg

ecast(Con

tent

Delivery)

Peer-to-Peerdistribu

tion

“Peer-to-peersystem

saredistribu

ted

system

sconsistin

gof

intercon

nected

nodesable

toself-organizeinto

network

topo

logies

with

thepu

rposeof

sharing

resourcessuch

ascontent,[...]

with

out

requ

iringtheinterm

ediatio

nor

supp

ort

ofaglob

alcentralized

server

orauthority.”

(And

routsellis-Theotok

isandSpinellis20

04)

-client

toclient

commun

ication

-And

routsellis-Theotok

isandSpinellis(200

4)-BitT

orrent

(DNA)

-Internetbased

-DeBoever(200

7)-Sharm

anNetworks

(KaZ

aA)

-Kwok

etal.(200

2)-Joo

stN.V.(joo

st)

-Kon

tiki(EnterpriseVideo

Delivery)

6 J. Wulf, R. Zarnekow

Table 1. In this section, a short overview of each technologyis given.

The Internet was originally designed to support highlyrobust communication. The first commercial World WideWeb services supported the retrieval of textual HTTPpages. Data was exchanged between centralized serversand distributed clients via the IP protocol based on the besteffort principle (Centralized Internet Hosting, Leighton2009; Tanenbaum 2003, pg. 4, Underwood 2001). Today,many information services are still hosted centrally.Tchibo.de, for example, a popular German online store, ishosted by the managed services provider Easynet (Easynet2009). Early Internet services did not impose high QoSrequirements. But since the original development of theOpen Systems Interconnection (OSI) basic reference model(ISO 1989), the requirements of information servicesdistributed via the Internet have changed radically (Faratin2007). Services with high real-time requirements aredisturbed if transported across large network distanceswhere packets have to traverse multiple network intercon-nection points and packet delay exceeds tolerable bounds.Content providers in the Internet are confronted withextreme demand peaks (such as flash crowds, Pathan andBuyya 2008) because the Internet has become a massmedium. Data intensive services have emerged due togrowing access capacity, which further increases the risk oftransportation overloads.

Two different approaches have emerged to supportinformation services which cannot satisfactorily be realizedthrough centralized hosting and best-effort transmission(Faratin 2007; Leighton 2009; Xiao 2008): the introductionof a QoS service model (as applied in direct homing), andthe installation of Internet overlay structures (i.e. CDNs orP2P networks). The goal of a QoS service model is toenable a QoS which is superior to the best-effort QoS in theInternet. This is realized by a modification of the traditionalInternet architecture: technologies, which allow packetdifferentiation or capacity reservation, must be introduced(Xiao and Ni 1999; Zhao et al. 2000; Huston 2000, pg.399). A QoS service model can either be realized on adedicated infrastructure, or by modifying the presentarchitecture of the public Internet. Presently, the QoSservice model is mostly realized in proprietary networks.For this reason, content distribution via the QoS servicemodel is only realized in close cooperation between contentand network providers: the content is directly homed on theQoS capable network, which provides a direct customeraccess to the content provider (direct homing). Well knownexamples in practice are the IP TV offerings of networkproviders, such as AT&T (U-Verse), BT (BT Vision) orDeutsche Telekom (T-Entertain).

In contrast, Internet overlay structures do not alter thetraditional Internet architecture, but provide value added

functionality on the application layer: An Overlay is a set ofservers deployed across the Internet that [...] in some waytake responsibility for the forwarding and handling ofapplication data in ways that are different from or incompetition with what is part of the basic Internet (Clark etal. 2005). Two overlay technologies for content delivery haveemerged which are specifically designed to optimize QoS anddelivery costs of IP distribution: CDNs and P2P networks, aspointed out by Faratin (2007) and Saroiu et al. (2002).

CDN providers operate a network of servers which areplaced in multiple autonomous systems and strategicallydistributed across the Internet (Dilley et al. 2002; Pallis andVakali 2006; Pathan and Buyya 2008; Vakali and Pallis2003). In order to improve QoS performance, e.g. transmis-sion delay and the effective bandwidth, objects such as largefiles or multimedia content are strategically distributedacross these servers. In addition to caching, CDN providersapply technologies such as prefetching, route optimization,and sophisticated request routing mechanisms. As anexample, France Televisions implements its Video onDemand offering via the content delivery network of theCDN provider Akamai (Akamai 2008).

P2P distribution systems (Androutsellis-Theotokis andSpinellis 2004; De Boever 2007; Kwok et al. 2002) aredesigned to make use of resources provided by endcustomers, such as storage space and processing power, inorder to establish a distributed storage medium throughwhich information service providers are able to distributetheir content. In addition to content replication and caching,P2P systems provide functionalities, such as the distributedpositioning and routing of objects, secure storage, accesscontrol, and authentication. Blizzard Entertainment, forexample, partners with BitTorrent Inc. to distribute thesoftware for the game World of Warcraft via P2Ptechnology (Blizzard 2009). In contrast to Pathan andBuyya (2008) and in line with Leighton (2009) and Saroiuet al. (2002), CDNs are distinguished from P2P networks:whereas CDN providers centrally manage a network ofsurrogate servers, P2P networks are established through thecontributions of independent peers.

In the following, a set of information service character-istics is presented which potentially bear a relationship todistribution performance. These relationships are thentested in the subsequent section.

Determinates of distribution quality

The quality of service distribution can be measured directlythrough the delay, error tolerance and effective bandwidth(equates to data load per second), provided by the distributionmechanism. Requirements of information services on thedistribution quality are taken from the CoS conceptualizations

Technologies for the Electronic Distribution of Information Services 7

(Marchese 2007, pp. 5–8). Information service providers are,nevertheless, unable to estimate the potential quality of adistribution technology prior to implementation. Moreover,the end-to-end measurement of distribution quality for arepresentative set of customers is a very complex task due tothe constitutive service characteristics, as explained above.

Information services expose heterogeneous require-ments, not only on the quality of service distribution, butalso on its implementation. The way information servicesare created and consumed potentially has a direct impact onthe performance of a specific distribution technology. Thenovel approach of the following analysis is the identifica-tion of dependencies of the quality provided by a servicedistribution technology based on the characteristics ofservice production and consumption (cf. Fig. 3, Table 2).These potential characteristics were identified in expertinterviews and through an analysis of the literature ondistribution technologies as presented above.

Relevant characteristics of service production are thedata load per service execution, the centrality of informa-tion origination, the interactivity of communication, thesimultaneity of provisioning and demand, and the servicerevenues. Data load per service execution describes howmuch traffic load a single service execution generates.Traffic load, being a main cost factor, is an important aspectin content distribution and influences technological deci-sions such as content caching strategies and serverlocalization. Centrality of information origination describeshow data originates and is fed into the network. Centralinformation origination implies that content originates at acentral location and, as such, can be distributed from acentral server. In contrast, distributed information origina-tion implies that content is produced at distributedlocations, e.g. the end customers’ clients. Content distribu-tion must, in this case, be carried out from distributedlocations, which leads to different prerequisites, e.g.concerning the application of route optimization strategies.Interactivity of communication characterizes the traffic flowbetween communication end points: in non-interactive

communication, a request by a client is typically followedby a large downstream traffic flow from a server. In interactivecommunication, traffic flows between end points have similarcharacteristics. Interactivity of communication plays animportant role in content distribution, e.g. regarding theimportance of route optimization technologies and theefficiency of content prefetching. Real-time services arecharacterized by a simultaneity of the provisioning anddemand for this content. As there exists no time for caching,such services rely mostly on route optimization strategies. Incontrast, elastic services are characterized by a time gapbetween information provisioning and demand, and for thisreason allow the application of buffering and cachingtechnologies. Service revenues, i.e. the revenues a serviceprovider generates per service execution, potentially provideinformation about the willingness of a service provider toinvest in high distribution quality. Distribution technologiesvary strongly with regard to the capital and operationalexpenses they induce. Low revenues do not allow highinvestments in service distribution and signal an endcustomer’s tolerance for low distribution quality. In contrast,high revenues provide information service providers withmore flexibility to invest in high distribution quality.

The relevant characteristics of service consumption arethe customer segment, demand locality, level of demand,consumer mobility, consumption frequency, and securitylevel. The customer segment, e.g. business or privatecustomers, potentially has an impact on the servicedistribution: business-to-business services in general im-pose stricter requirements on distribution quality, e.g.regarding service availability and reliability, than business-to-consumer services. The locality of demand also has asignificant impact: in order to meet a global servicedemand, comprehensive distribution technology implemen-tations need to be applied, which often require higherinvestments than meeting a demand that is restricted to asmall, well defined geographical area. The level of demanddescribes whether a service is considered a niche or a massservice. A niche service imposes different requirements for

Fig. 3 Characteristics of information services influencing the quality of service distribution

8 J. Wulf, R. Zarnekow

Tab

le2

Criteriacatego

ries

Criterion

Class

(Num

berof

Observatio

ns)

Mean

value

Stand

ard

deviation

12

34

5

Centrality

ofInform

ation

Originatio

ndecentral(37)

central(66)

//

/1.64

0.48

Con

sumer

Mob

ility

static

(21)

distribu

tedstationary

access

(59)

onthemov

e(23)

//

2.02

0.66

Customer

Segment

pure

B2C

(71)

nospecific

focus(20)

pure

B2B

(12)

//

1.43

0.69

DataLoadperSecon

d<64

kbitp

s(0)

[64kb

itps—

384kb

itps)

(17)

[384

kbitp

s—1Mbp

s)(44)

[1Mbp

s—4Mbp

s)(35)

>=4Mbp

s(7)

3.31

0.83

DataLoadperService

Executio

n<1

MB(0)

>=1

MB(21)

>=15

MB

(55)

>=50

0MB(20)

>=5

GB

(7)

3.13

0.81

Delay

Tolerance

<=10

0ms(24)

(100

ms—

1s]

(11)

(1s–4

s](36)

(4s–6

s](9)

>6

s(23)

2.96

1.43

Dem

andLocality

region

al(4)

coun

try(36)

continent(16)

glob

al(47)

/3.03

0.98

Error

Tolerance

0%(68)

<=0,3%

(0)

<=1%

(19)

>1%

(16)

/1.83

1.21

Frequ

ency

ofCon

sumption

>=1peryear

(15)

>=1permon

th(21)

>=1perweek(14)

>=1perday(36)

>=1perho

ur(17)

3.18

1.33

Interactivity

ofCom

mun

ication

one-way

commun

ication(65)

two-way

commun

ication

(con

trol

andcontentdata

flow

s)(13)

two-way

commun

ication

(hom

ogeneous

data

flow

s)(25)

//

1.61

0.85

Level

ofDem

and

nicheservice(19)

regu

larservice(37)

massservice(47)

//

2.27

0.76

Security

Level

noncritical

data

(41)

data

allowingendcustom

erprivacyintrusion(34)

strategicbu

siness

data

(20)

data

criticalwith

regard

tolegalandregu

latory

issues

(8)

/1.95

0.95

Service

Revenues(inUS$)

none

(2)

(0–1

0)(48)

[10–10

0)(35)

[100

–100

0)(10)

>=10

00(8)

2.75

0.95

Sim

ultaneity

ofProvision

ing

andDem

and

asyn

chrono

us(60)

quasisynchron

ous

(delay

tolerant)(8)

real-tim

e(35)

//

1.76

0.93

Technologies for the Electronic Distribution of Information Services 9

content delivery than a mass customer service. This affects theefficiency of technologies such as caching and capacityreservation. Consumer mobility also influences servicedelivery: In the case where a service is solely consumedfrom a single access line, such an access line can beexplicitly configured prior to service distribution. In contrast,mobile services do not allow a pre-configuration of accesslines, because such services must be available on multipleaccess lines, which cannot be determined in advance. Therealso exists a relationship between the frequency of serviceconsumption and the efficiency of content distributiontechnologies: the more regularly a service is consumed,the more efficient is application of refined distributiontechnologies, such as the installation of local caches. Thesecurity level of an information service describes therequirements of a service customer regarding the securityof data exchange. Sensitive customer data needs to beprotected more strictly than information, which is open to thepublic and does not carry any security risks. This affects, forexample, the employment of public infrastructure.

Value proposition analysis

103 IP based services were analyzed in order to identifydifferences in the value propositions of the four distributiontechnologies presented above. The services reveal insightsinto the value propositions of the distribution technologiesthey rely on.

Research methodology

The analysis was carried out according to the directives forcontent analysis research (Kassarjian 1977; Kolbe andBurnett 1991).1 Adequate sampling was ensured by pre-analyzing reference customer lists and use cases ofdistribution providers before making representative sampleselections. The distribution methods of the samples wereverified by employing various network tools such astraceroute and WHOIS.

The selected services were classified with respect to thedirect distribution quality criteria and the characteristics ofservice production and consumption presented in theprevious section. Table 2 contains a listing of the categoriesfor each criteria.

In order to guarantee data validity, the classificationwas carried out independently by five IT service experts,who were carefully chosen with respect to three selectioncriteria: education, work experience in the IT servicesfield, and expertise in IT service application, development,

and management. These selection criteria ensured that thejudges were familiar with the analyzed informationservices and their underlying technologies, and ultimatelyguaranteed the judges’ capability to produce accurateclassifications.2

Prior to coding, the judges were familiarized with thecriteria and category definitions. Judges were asked to carryout the following process steps during categorization:service and business model research, service testing, andresearch on service usage. Service and business modelresearch comprised the retrieval of information about theservice type, service technologies, target customers andunderlying business models from the service provider. Thisprocess step supplied valuable information for criteria suchas the customer segment, service revenues and securitylevel. The definition of a service type, based on CoSconceptualizations, supported the determination of delaytolerance, error tolerance, and data load per second for aspecific service. In a second step, judges used and tested theservices where possible. In order to track technical serviceparameters, such as data loads, judges were providedwith network and communication monitoring tools. Incombination with their personal usage experience withservices of similar types, testing gave valuable insightsfor criteria, such as the simultaneity of provisioning anddemand, centrality of information origination and con-sumer mobility. In a third step, judges were asked to gatherinformation on the usage and diffusion of a service frominformation technology research companies and web trafficmonitoring providers. This information was required for thecategorization of criteria such as the level of demand anddemand locality. After having followed these process steps,judges were asked to categorize an information service forall predefined criteria, and to leave blanks where a clearcategory distinction could not be detected. After collectingthe results from all judges, abnormal value discrepancieswere jointly analyzed and, in case of obvious misclassifi-cations, eliminated in a post valuation round. The fiveclassifications were merged in a final step by calculatingmean values.

Discriminant analyses (Klecka 1980) were carried out onthis data. The results indicate the relative importance of acriterion for the choice of a distribution method. For eachdistribution method, the group of services supported by thisspecific method was distinguished from the rest andinsights were gained into the discriminatory effect ofservice criteria. In each case, a univariate ANOVA and astepwise discriminant analysis were performed (Tables 3, 4,5 and 6). In the following section, all service criteria whichwere considered significant in both tests are discussed.

1 Refer to Table 8 in the Appendix for a summary of the directives andconformance justifications.

2 For an overview of the judges’ qualifications, readers are referred toTable 9 in the Appendix.

10 J. Wulf, R. Zarnekow

Table 3 Discriminatory service criteria for centralized hosting

Univariate analysis Stepwise discriminant analysisa,b,c

Group mean ‘centralizedhosting’ (n=27)

Group mean‘others’ (n=76)

F for group meanequality test

F toremove

F toenter

Wilks-Lambda (removal& entry criterion)

Discriminantloadingsd

DemandLocality 2.259 3.303 28.365*** 31.580*** .793 .651

DataLoadperExec 2.630 3.303 15.623*** 12.237*** .676 .483

ErrorTolerance 1.148 2.079 13.314*** 11.106*** .669 .446

ConsumerMobility 2.1111 1.9868 .712 3.312 .582 .169

CentralityOfInfoOrg 1.5556 1.6711 1.145 3.197 .582 −.128SimultanProvDem 1.3704 1.8947 6.630** 3.155 .583 .051

ServiceRevenues 2.3333 2.8947 7.447*** 2.540 .586 .088

LevelOfDemand 2.0370 2.3553 3.613* 2.537 .586 −.016DelayTolerance 2.8889 2.9868 .093 1.720 .591 −.119SecurityLevel 2.2222 1.8553 3.008* 1.335 .593 −.033DataLoadPerSec 3.0741 3.3947 3.042* 1.071 .595 .105

FrequencyOfCons 3.1111 3.2105 .110 .434 .599 .131

CustomerSegment 1.2593 1.4868 2.163 .176 .600 .245

InteractivityOfCom 1.4074 1.6842 2.114 .029 .601 .153

***p<0.01, **p<0.05, *p<0.10aMinimal partial F-statistic for acceptance: 3.84, Maximal partial F-statistic for exclusion: 2.71bWilks Lambda of Discriminant Function: 0.601, Number of Steps: 3c Class mean values of discriminant function: Centralized Hosting=-1.35, others=0.48d Correlation between discriminating variables and the canonical discriminant function

Table 4 Discriminatory service criteria for direct homing

Univariate analysis Stepwise discriminant analysisa,b,c

Group mean ‘directhoming’ (n=20)

Group mean‘others’ (n=83)

F for group meanequality test

F toremove

F toenter

Wilks-Lambda (removal& entry criterion)

Discriminantloadingsd

ConsumerMobility 1.000 2.265 143.381*** 400.532*** .385 −.336ErrorTolerance 2.900 1.578 23.688*** 104.456*** .155 .137

DataLoadPerSec 3.250 3.325 .132 33.869*** .100 −.010InteractivityOfCom 2.600 1.373 48.757*** 9.182*** .081 .196

ServiceRevenues 3.850 2.482 49.721*** 7.588*** .080 .198

FrequencyOfCons 4.500 2.867 31.314*** 4.915** .078 .157

DelayTolerance 1.150 3.398 65.040*** 4.061** .077 −.227SimultanProvDem 2.800 1.506 43.997*** 1.875 .072 .266

CustomerSegment 2.100 1.265 29.887*** .095 .074 .179

CentralityOfInfoOrg 1.200 1.747 25.786*** .898 .073 −.083DataLoadperExec 3.800 2.964 20.289*** .026 .074 .116

SecurityLevel 2.700 1.771 17.813*** .360 .074 .164

LevelOfDemand 1.900 2.361 6.307** .371 .074 −.020DemandLocality 2.700 3.108 2.822* 1.585 .073 .066

***p<0.01, **p<0.05, *p<0.10aMinimal partial F-statistic for acceptance: 3.84, Maximal partial F-statistic for exclusion: 2.71bWilks Lambda of Discriminant Function: 0.074, Number of Steps: 7c Class mean values of discriminant function: DH=7.14, others=−1.72d Correlation between discriminating variables and the canonical discriminant function

Technologies for the Electronic Distribution of Information Services 11

Table 5 Discriminatory service criteria for CDNs

Univariate analysis Stepwise discriminant analysisa,b,c

Group mean‘CDN’ (n=30)

Group mean‘others’ (n=73)

F for group meanequality test

F toremove

F toenter

Wilks-Lambda (removal& entry criterion)

Discriminantloadingsd

ConsumerMobility 2.400 1.861 16.264*** 24.990*** 0.796 0.530

ServiceRevenues 2.900 2.694 0.994 23.533*** 0.787 0.131

CentralityOfInfoOrg 1.900 1.528 14.212*** 12.356*** 0.714 0.495

LevelOfDemand 2.467 2.181 3.089* 5.241** 0.667 0.231

DataLoadperExec 3.133 3.139 0.001 3.211 0.613 0.257

DataLoadPerSec 3.467 3.250 1.441 2.712 0.616 0.382

InteractivityOfCom 1.267 1.764 7.604*** 1.517 0.623 −0.255ErrorTolerance 1.733 1.889 0.349 1.146 0.626 0.077

DelayTolerance 3.500 2.722 6.605** 1.050 0.626 0.209

DemandLocality 3.367 2.875 5.514** 0.949 0.627 0.149

FrequencyOfCons 2.800 3.347 3.621* 0.491 0.630 −0.154SimultanProvDem 1.633 1.819 0.837 0.338 0.631 −0.177CustomerSegment 1.300 1.486 1.519 0.254 0.631 −0.103SecurityLevel 1.833 2.014 0.757 0.017 0.633 −0.097

***p<0.01, **p<0.05, *p<0.10aMinimal partial F-statistic for acceptance: 3.84, Maximal partial F-statistic for exclusion: 2.71bWilks Lambda of Discriminant Function: 0.633, Number of Steps: 4c Class mean values of discriminant function: CDN=1.17, others=−0.49d Correlation between discriminating variables and the canonical discriminant function

Table 6 Discriminatory service criteria for P2P distribution

Univariate analysis Stepwise discriminant analysisa,b,c

Group mean‘P2P’ (n=26)

Group mean‘others’ (n=77)

F for group meanequality test

F toremove

F toenter

Wilks-Lambda (removal& entry criterion)

Discriminantloadingsd

DemandLocality 3.692 2.805 18.476*** 20.026*** 0.782 −0.584ServiceRevenues 2.154 2.948 15.639*** 8.677*** 0.708 0.537

SecurityLevel 1.231 2.195 24.426*** 4.224** 0.678 0.671

InteractivityOfCom 1.462 1.662 1.074 2.401 0.635 0.366

SimultanProvDem 1.500 1.844 2.684 1.557 0.640 0.400

DataLoadperExec 3.115 3.130 0.006 1.255 0.642 0.190

CustomerSegment 1.231 1.494 2.831* 1.081 0.643 0.350

ConsumerMobility 2.269 1.935 5.246** 1.053 0.644 −0.462DelayTolerance 3.808 2.675 13.761*** 0.609 0.647 −0.381DataLoadPerSec 3.423 3.273 0.637 0.596 0.647 0.019

ErrorTolerance 1.846 1.831 0.003 0.071 0.650 −0.052CentralityOfInfoOrg 1.769 1.597 2.505 0.028 0.650 −0.187FrequencyOfCons 2.692 3.351 4.916** 0.006 0.651 0.289

LevelOfDemand 2.577 2.169 5.928** 0.004 0.651 −0.322

***p<0.01, **p<0.05, *p<0.10aMinimal partial F-statistic for acceptance: 3.84, Maximal partial F-statistic for exclusion: 2.71bWilks Lambda of Discriminant Function: 0.651, Number of Steps: 3c Class mean values of discriminant function: P2P=−1.25, others=0.42d Correlation between discriminating variables and the canonical discriminant function

12 J. Wulf, R. Zarnekow

Centralized hosting

Centralized hosting technology represents the traditionaldistribution method within IP networks. Content is hostedon a single server which is connected to the network.Packets are routed through interconnected networks to therequesting client. QoS in pure centralized hosting distribu-tion varies highly, depending on the end customer’slocation and the specific route: thus QoS will most likelybe perfectly acceptable if the end customer is connected tothe same network service provider as the hosting provideror if traffic does not have to traverse large distances. QoSmay exceed acceptance thresholds if traffic is exchangedover large distances through various interconnected net-works. Both tests (Table 3) show that demand locality is themost significant criterion with respect to the choice forcentralized hosting distribution: the more demand is limitedto a specific region, the more effective centralized hostingbecomes as a distribution mechanism. This is well in linewith the argument above that interconnections and geo-graphical distance have a significant negative impact onQoS. The tests also reveal that centralized hosting isparticularly suitable for services with a low data load perservice execution. Hosting providers charge per stored andaccessed data volume; network service providers charge foraccess capacity. Moreover, data transport becomes moreexpensive if more interconnections are involved. Contentproviders, therefore, only choose centralized hosting forservices with low data volume. In addition, low errortolerance is well supported by centralized hosting. Thisresult suggests that highly error tolerant services, such asvideo conferencing, telephony, and multimedia streaming,are generally not provided through centralized hosting.

Direct homing

Within direct homing, IP traffic is not routed as part of thebest effort Internet class, but instead as a privileged class orthrough reserved capacities. Consumer mobility is thedominant selection criterion for direct homing (Table 4).This is caused by the fact that direct homing only supportsservice consumption with guaranteed end-to-end QoS frompre-configured access lines. Thus, end customers can notmake use of reserved capacities from arbitrary accesspoints, but only from dedicated access points. Hence, thistechnology is not adequate for mobile end customerservices. IP TV with guaranteed QoS, for example, isinstalled with a configured router on a single access line. Ahighly reliable and configurable QoS with low latency isaspired through the privileged treatment of direct homingtraffic. This is especially valuable for services with a highdegree of interactivity, such as Voice over IP (VoIP). Thetests affirm that directly homed services are indeed

characterized by a high level of interactivity and errortolerance. As data transport is carried out on predeterminedroutes, interactive services can be supported more effectivelythan by other distribution technologies. The high errortolerance of directly homed services suggests that directhoming is often used for error tolerant services such as voiceand streaming services. Traffic differentiation or routingthrough dedicated networks does not necessarily imply theneed to install separate hardware, but requires at leastadditional investments in the installation and operation ofrouting systems. As direct homing is often more cost intensivethan other distribution methods, direct homing is onlyadequate for services with a high frequency of consumptionand for services which generate high revenues. Otherwise, thehigh investments are not justifiable. In contrast to intuition,direct homing does not primarily address a B2B servicecontext. Direct homing represents the only distributiontechnology, where delay tolerance is a significant criterion.In contrast to the other distribution methods, where the directdiscriminatory effect of delay tolerance is insignificant, thisaspect is directly taken into account for direct homing.

Content delivery networks

CDN providers operate a complex network of edge servers.Content is distributed from and cached on these serversbased on strategies which take into account the location ofend customers, as well as network and server character-istics. CDNs, in contrast to direct homing, support a highcustomer mobility during service consumption (Table 5).CDN service coverage is not limited to configured accesslines and networks. CDNs support mobile services and arewell suited to deliver services such as video streaming,which are consumed en-route. CDN technology relies on asingle central origin server, that contains core serviceinformation and that directs content or requests to the edgeservers. On the other hand, CDN is not well suited forservices without a central origination of information, such ascommunication services in which each end customer feedsdata into the network. For such services, content caching onthe edge servers can not be carried out efficiently. Contentcaching becomes more efficient with higher cache hit ratios,i.e. the ratio of cached versus total documents requested. Forthis reason, CDNs particularly address services with a masscustomer focus. For niche services, cache hit ratios tend tobe low, which decreases the efficiency of cache networks.As the content needs to be transported to the edge serversprior to its distribution to end customers, this distributionmethod does not seem adequate for services where theorigination of information coincides with the demand forthis information. However, this hypothesis can not beconfirmed. CDN providers constantly improve their capa-bilities to support real-time services, e.g. through the

Technologies for the Electronic Distribution of Information Services 13

development of route optimization technologies. For quasireal-time services such as live streaming, CDN technologyis widely applied (Pathan and Buyya 2008).

P2P distribution

P2P distribution does not require large capital expenses.The main resource is P2P software running on endcustomers’ clients with established network connections.Content providers do not need to constantly feed contentinto the network from central servers and therefore savetransit costs. As overall distribution costs increase with theglobal nature of demand, P2P especially addresses serviceswith a global demand (Table 6). This is the most importantcriterion for choosing P2P distribution. In the case of astrictly local service demand, services could be distributedvia centralized hosting without significant distributioncosts. In case of a global demand, content distribution viaP2P incurs significantly less transit costs. As contentproviders do not have to feed content into the network foreach customer request, this distribution method is particu-larly attractive for services which generate low revenuesand do not justify investments in a costly distributiontechnology with high QoS guarantees. Hence, cost savingsin distribution are a dominant decision factor. More than inother distribution methods, content is exposed to the risk ofmanipulation. Since content is cached by third party clientsprior to its distribution, the risk of unauthorized serviceaccess or service modification is high compared to otherdistribution methods. For services with high securityrequirements (e.g. file storage and exchange services forbusinesses), this risk is generally not tolerable.

Summary

In this article, the quality provided by different technologiesfor information service distribution is discussed. Ananalysis of 103 information services was carried out inorder to identify the main characteristics influencing aninformation service providers choice of a specific distribu-

tion method. An underlying assumption of this approach isthat the information services analyzed are best supported bytheir associated distribution methods and, as such, allowdeductions on the value proposition of the distributiontechnologies. This assumption is justified by the rationalethat firms seek an optimal distribution choice. The results ofour analysis reveal that distribution technologies varystrongly with respect to the information services theysupport. A set of characteristics concerning service produc-tion and consumption was identified which have a stronginfluence on the quality provided by distribution technolo-gies. Distribution technologies and their significant servicecharacteristics are summarized in Table 7.

From the set of hypotheses, i.e. the set of characteristicswhich were taken into account, several were provenirrelevant: most notably, criteria directly describing thequality of service distribution, namely packet delay anddata load per second, have an insignificant influence on thechoice of a distribution technology. This means that there isno distribution technology which generally provides asuperior distribution quality. Instead, information serviceproviders need to match their individual service character-istics to the technological capabilities of distributionmethods for an optimal choice. Moreover, the simultaneityof provisioning and demand and the customer segment areirrelevant service criteria. From this it follows that theanalyzed distribution technologies all have real-time capa-bilities and that none is specifically tailored for a businessor an end consumer service scenario.

This work represents an initial analysis comparing thecapabilities of service distribution technologies. Theresults provide information service providers with prac-tical guidance in their choice of a distribution provider.They allow the selection of a distribution method basedon a matching of service characteristics and capabilitiesof distribution technologies.

Distribution technologies are constantly being enhanced.CDN and P2P technologies, for example, are most recentlybeing applied jointly to reach a more efficient servicedistribution. Blizzard Entertainment uses AKAMAI’s CDNand BitTorrent’s P2P technology to distribute its World of

Table 7 Distribution technologies and their significant service characteristics

Distributiontechnology

Significant service characteristics

Centralized Hosting Local demand, low data load per service execution, low error tolerance

Direct Homing Low consumer mobility, high error tolerance, high interactivity of communication,high service revenues, high frequency of consumption, low delay tolerance

Content Delivery Networks High consumer mobility, high centrality of information origination, high level of demand

P2P Distribution Global demand, low service revenues, low security level

14 J. Wulf, R. Zarnekow

Warcraft Software (Blizzard 2009). The future evolution ofinformation service distribution is dependent on deeperinsights into the quality of information services and theinfluence of distribution performance. Zeithaml et al. (2000)state that a negative perception of information service qualityresult from several gaps, two of which are critical for servicedistribution: the information gap represents differencesbetween customer requirements and a service provider’sbeliefs about requirements. The design gap describesdiscrepancies between the identified requirements and theimplemented service. By filling the information gap, serviceproviders gain a more profound understanding of the

constitutive characteristics of service quality. This knowledgeis a prerequisite for service implementation. The design gapis closed by correctly matching requirements with servicespecifications and by achieving these specifications duringservice operation. In this phase, the required quality ofservice distribution for specific information services isidentified. This work represents an approach to close thesegaps by establishing a relationship between service require-ments and distribution methods which are currently inoperation. A clear understanding of both gaps is aprerequisite for the further development and effectiveapplication of distribution technologies.

Appendix

Table 8 Conformance with content-analysis research directives (based on Kolbe and Burnett 1991)

Objectivity Definition of rules and procedures Reporting of categories and definitions

Judge training Presentation of category definition prior to coding

Unit measure pretesting Pretesting conducted by authors

Judge independence Engagement of four non-authoring judgesMultiple judges

Systematization Prohibition of bias between categoryselection and thesis

Avoidance of result predetermination by definitionof categories in cooperation with independent expertsand based on literature

Hypothesis testing Hypothesis formulation and foundation

Sampling Methods Generalizability of sample Selection of representative published use cases

Manageability of sample size Sample size of 103 applications

Reliability Categorical reliability Clear definition of categories

Interjudge reliability Introduction of feedback cycle to avoidinterjudge disagreements

Table 9 Information on judge qualifications

Judge Education Work experience inthe IT services field

Expertise in IT service application, development,and management

1 Information Engineering and Management (MSc) 6 years Application Programming, Software Engineering,Network Technologies, IT Service Management

2 Industrial Engineering and Management (MSc) 3 years Web Design, IT Service Quality Management,Internet Technologies

3 Information Systems Management (MSc) 8 years Application Programming, Software Engineering,IT Service Management

4 Industrial Engineering and Management (BSc) 1 ½ years Web Design, IT Service Management,Internet Technologies

5 Computer Sciences (BSc) 3 years Application Programming, Network Technologies,Server Configuration

Technologies for the Electronic Distribution of Information Services 15

Tab

le10

Listof

analyzed

inform

ationservices

Nam

eof

inform

ationservice

Inform

ationserviceprov

ider

Typ

edescription

Distributiontechno

logy

Distributionprov

ider

ICQ

/AOLInstantMesseng

erAOL

InstantMesseng

er,Internet

Telepho

nyP2P

AOL

20th

Century

Fox

Film

s20

thCentury

Fox

Video

Streaming

P2P

Vuze

301Records

Music

Album

301Records

Music

Dow

nload

P2P

Kazaa

ABS-CBN

TV

ABS-CBN

Global

TV

Streaming

CDN

Edg

eStream

Aeria

Gam

ePatches

Aeria

Gam

esGam

eDistribution

P2P

BitT

orent

Altn

etMusical

Dow

nloadService

Altn

etMusic

Dow

nload

P2P

Kazaa

amazon

.deWeb

Sho

pAmazon

OnlineSho

pIP

Transit

Level3

AMD

Drivers

AMD

DriverDow

nload

CDN

Akamai

Aud

iWebsite

Aud

iAG

BusinessInternet

Representation

CDN

Akamai

B2B

-trade.net

B2B

-Trade

Ltd.&

Co.

KG

B2B

Marketplace

IPTransit

Hosteurop

e

morgenp

ost.d

eBerlin

erMorgenp

ost

New

sPortal

IPTransit

Arcor

bild.de

BILD

digitalGmbH

&Co.

KG

New

sPortal

IPTransit

ColtTelecom

mun

ications

billiger.d

ebilliger.d

eInform

ationAgg

regator

IPTransit

IPExchang

eGmbH

Bluew

inTV

Swisscom

TV,Video

Streaming

DirectHom

ing

Swisscom

Bollywoo

dMov

ies

IndiaFM.com

Mov

ieDow

nload

P2P

Kazaa

Music

Video

Streaming

Son

yBMG

TV,Video

Streaming

DirectHom

ing

BT

BusinessCon

nect

Professional

Swisscom

VoIP

DirectHom

ing

Swisscom

BusinessVideo

1000

MXP

T-Systems

Video

Con

ferencing

DirectHom

ing

T-Systems

CathayPacific

AirwaysWebsite

CathayPacific

Airways

Website

CDN

Akamai

CNBCHom

epage

CNBC

New

sPortal

CDN

Akamai

Com

edyCentral

Series

Com

edyCentral

Video

Streaming

P2P

BitT

orent

tagesspiegel.de

Der

Tagesspiegel

New

sPortal

IPTransit

IPExchang

eGmbH

DirecTV

VoD

DirecTV

TV

Streaming

CDN

Techn

icolor

EDS

Dream

Works

Video

Clip

sDream

Works

Clip

Stream

CDN

Lim

elight

Networks

Ebay.de

Transactio

nPlatform

eBay

OnlineAuctio

nHou

seIP

Transit

Level3

ElectronicArtsGam

esSoftware

ElectronicArts

SoftwareDow

nload

CDN

Lim

elight

Networks

End

Customer

VoIP

Private

End

Customers

VoIP

DirectHom

ing

Telefon

icaDeutschland

GmBH

End

Customer

VoIP

Private

End

Customers

VoIP

DirectHom

ing

o2

End

Customer

VoIP

Private

End

Customers

VoIP

DirectHom

ing

Arcor

End

Customer

VoIP

Private

End

Customers

VoIP

DirectHom

ing

HanseNet

End

Customer

VoIP

Private

End

Customers

VoIP

DirectHom

ing

1&1

Equ

antManaged

IP-V

PN

Equ

ant

IPVPN

DirectHom

ing

Equ

ant

ESA

Portal

ESA

Clip

Stream

CDN

Akamai

FAZ.net

F.A.Z.ElectronicMedia

GmbH

New

sPortal

IPTransit

Versatel

ftd.de

Financial

Tim

esDeutschland

New

sPortal

IPTransit

Gruner+

Jahr

AG

&Co

16 J. Wulf, R. Zarnekow

Tab

le10

(con

tinued)

Nam

eof

inform

ationservice

Inform

ationserviceprov

ider

Type

description

Distributiontechno

logy

Distributionprov

ider

FranceTelevisions

VoD

FranceTelevisions

Video

onDem

and

CDN

Akamai

Gam

eShado

wSoftwareDow

nload

Gam

eShado

wSoftwareDow

nload

CDN

Lim

elight

Networks

Geizkragen.de

Geizkragen

Inform

ationAgg

regator

IPTransit

DTSSysteme

Goo

gleApp

sGoo

gleInc.

SaaS

IPTransit

Goo

gle

Goo

gleTalk

Goo

gle

Instantmesseng

er,Internet

Telepho

nyP2P

Goo

gle

guenstiger.de

guenstiger.de

Inform

ationAgg

regator

IPTransit

Inet

PeopleHostm

aster

hand

elsblatt.com

Handelsblatt

New

sPortal

IPTransit

circ

ITGMbH

HostelWebsite

HostelDavid

HostelWeb

Representation

IPTransit

StratoAG

IAHGam

ePatches

Infocomm

AsiaHolding

sPte

Ltd

(IAHGam

es)

SoftwareDow

nload

P2P

BitT

orent

IntelDriverDow

nload

Intel

SoftwareDow

nload

CDN

Akamai

InterCon

tinentalWebsite

InterCon

tinentalHotel

Group

Website

CDN

Akamai

IntraSelectVPN

T-Systems

IPVPN

DirectHom

ing

T-Systems

L’Equ

ipeHom

epage

L’Equ

ipe

New

sPortal

CDN

Akamai

Lionsgate

Film

sLionsgate

Video

Dow

nload

P2P

BitT

orent

MachinimaFilm

sMachinima

Mov

ieandClip

Dow

nload

P2P

Vuze

Managed

IP-V

PN

GlobalCrossing

IPVPN

DirectHom

ing

GlobalCrossing

Managed

IP-V

PN

TFM

Networks

IPVPN

DirectHom

ing

TFM

Networks

Managed

Voice

over

IPNetworks

T-Systems

VoIP

DirectHom

ing

T-Systems

Maxdo

meVoD

ProSiebenS

at.1

Media

&UnitedInternet

Video

onDem

and

CDN

Akamai

Ministryof

Sou

ndTV

Ministryof

Sou

ndMusic

P2P

Vuze

MSNBC

Webcast

MSNBC

Clip

Stream

CDN

Lim

elight

Networks

Music

Album

“The

Morning

Benders”

+1Records

Music

Dow

nload

P2P

Lim

eWire

Myd

eoEnterpriseMedia

Delivery

Myd

eoClip

Stream

CDN

Lim

elight

Networks

MyS

pace

Website

MyS

pace

SocialNetwork

CDN

Lim

elight

Networks

neu.de

Website

Neu.deGmbH

Partnerbö

rse

IPTransit

ColtTelecom

mun

ications

NXTbo

okMedia

OnlineBrochures

NXTbo

okMedia

printmaterialsforweb

use

CDN

MirrorIm

age

NY

PostHom

epage

NY

Post

New

sCDN

Akamai

OrvisWeb

Sho

pOrvis

OnlineStore

CDN

MirrorIm

age

PanTerra

On-Dem

andPlatform

SaaS

SUTHERLAND

NETWORKS

SaaS

DirectHom

ing

PanterraNetworks

Param

ount

PicturesVoD

Param

ount

Video

Dow

nload

P2P

BitT

orent

PetStore

Web

Sho

pPetStore

OnlineStore

CDN

MirrorIm

age

RajshriMedia

VoD

RajshriMedia

Video

onDem

and

CDN

Lim

elight

Networks

SageClubWebsite

SageClub

ClubHom

epage

IPTransit

NeueMedienMuenn

ichGmbH

SAPformedium

sizedbu

sinesses

AlliedPanelsEntwicklun

gs-und

Produ

ktions

GmbH

SaaS

DirectHom

ing

Freud

enberg

IT

SchuelerV

ZWebsite

Verlagsgrup

peGeorg

von

Holtzbrinck

GmbH

SocialNetwork

CDN

Panther

Exp

ress

Secon

dLife

LindenLabs

SoftwareDow

nload

IPTransit

Amazon

Technologies for the Electronic Distribution of Information Services 17

Tab

le10

(con

tinued)

Nam

eof

inform

ationservice

Inform

ationserviceprov

ider

Typ

edescription

Distributiontechno

logy

Distributionprov

ider

SegaSoftware

Sega

Gam

eDistribution

P2P

BitT

orent

shop

wahl.d

eWebsite

LeG

uide.com

Group

Sho

ppingDirectory

IPTransit

ColtTelecom

mun

ications

Sho

wtim

eFilm

sSho

wtim

eMov

ie,Series

P2P

Vuze

SiemensA&D

Software

SiemensA&D

SoftwareDow

nload

CDN

Akamai

Sipgate

Internet

Teleph

ony

Sipgate

Internet

Telepho

nyP2P

Sipgate

Sky

peInternet

Telepho

nyeB

ayInstantmesseng

er,Internet

Telepho

ny,Filesharing

P2P

eBay

Son

yEricssonSoftware

Son

yEricsson

SoftwareDow

nload

CDN

Akamai

Son

yPicturesFilm

sSon

yPictures

Mov

ieP2P

Vuze

Spacenatio

nsOnlineGam

eSpacenatio

nsBrowsergam

eIP

Transit

IPX

ServerGmbH

SPD

Website

SPD

Website

IPTransit

PIRONETNDH

AG

Sup

erbo

wlCom

mercial

Clip

sIFILM

Video,TV,Gam

esCDN

Lim

elight

Networks

StudiVZWebsite

Verlagsgrup

peGeorg

von

Holtzbrinck

GmbH

SocialNetwork

CDN

Panther

Exp

ress

tchibo

.deWebsite

Tchibo

OnlineSho

pIP

Transit

Easyn

etGmbH

Ten

Mile

TideMusic

Album

Ten

Mile

Tide

Music

P2P

Kazaa

TheCrownOnlineGam

eTheCrown

Browsergam

eIP

Transit

manitu

hosting

T-ho

meEntertainmentVoD

DeutscheTelekom

TV,Video

Streaming

DirectHom

ing

DTA

G

Trium

phVoIP

Trium

phVoIP

DirectHom

ing

DTA

G

U-Verse

VoD

AT&T

TV,Video

Streaming

DirectHom

ing

AT&T

Videoload

VoD

DeutscheTelekom

Video

Streaming

IPTransit

DTA

G

WarnerBros.Film

sWarnerBros.

Video

Dow

nload

P2P

BitT

orent

welt.d

eWebsite

Die

Welt

New

sPortal

IPTransit

Arcor

Windo

wsLiveMesseng

erMicrosoft

Instantmesseng

er,Internet

Telepho

nyP2P

Microsoft

World

ofWarcraftSoftware

Blizzard

SoftwareDow

nload

CDN

Lim

elight

Networks

World

ofWarcraftSoftware

Blizzard

SoftwareDow

nload

P2P

BitT

orent

www.blog.de

Website

Mok

onoGmbH

BlogSite

IPTransit

IPExchang

eGmbH

Yahoo

Messeng

erYahoo

Instantmesseng

er,Internet

Telepho

ny,Filesharing

P2P

Yahoo

ZattooIP

TV

Zattoo

P2P

TV

P2P

Zattoo

ZDFMediathek

VoD

ZDF

Video

OnDem

and

CDN

Akamai

18 J. Wulf, R. Zarnekow

References

Akamai (2008). Case study—france television interactive. AkamaiTechnologies, Inc.Website. http://www.akamai.com/html/customers/case_study_francetv.html. Accessed 4 June 2009.

Akamai (2009). Akamai: The leader in web application acceleration andperformance management. Streaming media, Akamai Technologies,Inc. Website. http://www.akamai.com. Accessed 19 May 2009.

Amit, R., & Zott, C. (2001). Value creation in e-business. StrategicManagement Journal, 22, 493–520.

Androutsellis-Theotokis, S., & Spinellis, D. (2004). A survey of peer-to-peer content distribution technologies. ACM ComputingSurveys, 36(4), 335–371.

BitTorrent (2009). Reduce content delivery bills. BitTorrent, Inc.Website. http://www.bittorrent.com/dna?csrc=splash. Accessed 5June 2009.

Blizzard (2009). Blizzard Downloader F. A. Q. Blizzard Entertain-ment, Inc. Website. http://www.worldofwarcraft.com/info/faq/blizzarddownloader.html. Accessed 4 June 2009.

Bowersox, D. J., Closs, D. J., & Helferich, O. K. (1986). Logisticalmanagement: A systems integration of physical distribution,manufacturing support, and materials procurement. New York:Macmillan.

Clark, D. D., Lehr, W., Bauer, S., Faratin, P., Sami, R., Wroclawski, J.(2005). The growth of internet overlay networks: Implications forarchitecture, industry structure and policy. In The Proceedings of the34th Telecommunications Policy Research Conference (TPRC 05).Arlington, VA, USA.

De Boever, J. (2007). Peer-to-peer networks as a distribution andpublishing model. In Proceedings of the 11th International Confer-ence on Electronic Publishing (pp. 175–188). Vienna, Austria.

Demirkan, H., Kauffman, R., Vayghan, J., Fill, H., Karagiannis, D., &Maglio, P. (2008). Service-oriented technology and manage-ment: perspectives on research and practice for the coming de-cade. Electronic Commerce Research & Applications, 7(4), 356–376.

Dilley, J., Maggs, B., Parikh, J., Prokop, H., Sitaraman, R., & Weihl,B. (2002). Globally distributed content delivery. Internet Com-puting, IEEE, 6(5), 50–58.

Easynet (2009). Easynet—facts and figures. Easynet Group Inc.Website. http://www.easynet.com/gb/en/about/fastFacts.aspx?SecondaryNavID=52. Accessed 4 June 2009.

Faratin, P. (2007). Economics of overlay networks: An industrialorganization perspective on network economics. In Proceedingsof the Joint Workshop on The Economics of Networked Systemsand Incentive-Based Computing (NetEcon+IBC), in conjunctionwith ACM Conference on Electronic Commerce (end custom-er’07). San Diego, California, USA.

Frost & Sullivan (2008). World video content delivery networksmarket. Frost & Sullivan Market Engineering Report #N425-70.

Gozdecki, J., Jajszezyk, A., & Stankiewiez, R. (2003). Quality ofservice terminology in IP networks. IEEE CommunicationsMagazine, 41(3), 153–159.

Huston, G. (2000). Internet performance survival guide: QoSstrategies for multiservice networks. Inc: John Wiley & Sons.

ISO (1989). Open systems interconnection—Basic reference model.International Organization for Standardization. Ref. Nr. 7498.Geneva, Switzerland.

ISO (2005). Quality management systems—Fundamentals and vocab-ulary. Third edition. International Organization for Standardiza-tion. Ref. Nr. 9000:2005(E). Geneva, Switzerland.

Kassarjian, H. H. (1977). Content analysis in consumer research.Journal of Consumer Research, 4(6), 8–18.

Klecka, W. R. (1980). Discriminant analysis. Quantitative applica-tions in the social sciences series. Thousand Oaks: Sage.

Kolbe, R. H., & Burnett, M. S. (1991). Content-analysis research: anexamination of applications with directives for improving researchreliability and objectivity. Journal of Consumer Research, 18(2),243–250.

Kwok, S. H., Lang, K. R., & Tam, K. Y. (2002). Peer-to-peertechnology business and service models: risks and opportunities.Electronic Markets, 12(3), 175–183.

Leighton, T. (2009). Improving performance on the internet. Commun,ACM, 52(2), 44–51.

Liao, Z., & Cheung, M. T. (2008). Measuring consumer satisfaction ininternet banking: a core framework. Communications of theACM, 51(4), 47–51.

Liu, C., & Arnett, K. P. (2000). Exploring the factors associated withWeb site success in the context of electronic commerce.Information & Management, 38(1), 23.

Marchese, M. (2007). QoS over heterogeneous networks. Hoboken:John Wiley & Sons Inc.

McKinnon, A. (1988). Physical distribution. In J. N. Marshall (Ed.),Services and uneven development (pp. 133–159). Oxford: OxfordUniversity Press.

Onecle (2009). Sample contracts—master services agreement—AkamaiTechnologies Inc. Onecle Inc. Website. http://contracts.onecle.com/akamai/msa.shtml. Accessed 25 April 2009.

Pallis, G., & Vakali, A. (2006). Insight and perspectives for contentdelivery networks. Communications of the ACM, 49(1), 101–106.

Parasuraman, A., Zeithaml, V. A., & Berry, L. L. (1985). A conceptualmodel of service quality and its implications for future research.Journal of Marketing, 49(4), 41–50.

Pathan, M., & Buyya, R. (2008). A taxonomy of CDNs. In R. Buyya, M.Pathan, & A. Vakali (Eds.), Content delivery networks (pp. 33–77).Germany: Springer-Verlag.

Saroiu, S., Gummadi, K. P., Dunn, R. J., Gribble, S. D., Levy, H. M.(2002). An analysis of internet content delivery systems. InProceedings of the 5th Symposium on Operating Systems Designand Implementation, San Francisco, USA.

Shapiro, C., & Varian, H. R. (1999). Information rules: a strategicguide to the network economy. Harvard Business School Press,Boston, Mass.

Tanenbaum, A. S. (2003). Computer networks (4th ed.). Upper SaddleRiver: Prentice-Hall.

Timmers, P. (1998). Business models for electronic markets. Elec-tronic Markets, 8(2), 3–8.

Underwood, K. (2001). Web hosting: An overview. Gartner TechnicalReport.

Vakali, A., & Pallis, G. (2003). Content delivery networks: Status andtrends. IEEE Internet Computing, IEEE Computer Society,November–December, 68–74.

Xiao, X., & Ni, L. M. (1999). Internet QoS: A big picture. IEEENetwork, March–April, 8–18.

Xiao, X. P. (2008). Technical, commercial and regulatory challenges ofQoS. An Internet service model perspective. Amsterdam: MorganKaufmann.

Zeithaml, V. A., Parasuraman, A., & Malhotra, A. (2000). Aconceptual framework for understanding e-service quality:Implications for future research and managerial practice. Work-ing paper, Cambridge, MA: Marketing Science Institute.

Zeithaml, V. A., Parasuraman, A., & Malhotra, A. (2002). Servicequality delivery through web sites: a critical review of extantknowledge. Journal of the Academy of Marketing Science, 30(4),362–375.

Zhao, W., Olshefski, D., & Schulzrinne, H. (2000). Internet qualityof service: An overview. Working Paper, Columbia University,New York.

Technologies for the Electronic Distribution of Information Services 19