A Hedonic Price Index for Cloud Computing Services

Persefoni Mitropoulou, Evangelia Filiopoulou, Stavroula Tsaroucha,

Christos Michalakelis and

Mara Nikolaidou

Department of Informatics and Telematics, Harokopio University of Athens, 9 Omirou str, Tavros, Athens, Greece

Keywords: Cloud Computing, Infrastructure-as-a-Service, Pricing Models, Hedonic Price Indices.

Abstract: Cloud computing is an innovative business model, being developed at a fast pace during the last years, offering

many operational and economic benefits to both the demand and the supply side of the ICT market.

Infrastructure as a Service (IaaS), which includes control of fundamental computing resources, is expected to

be the fastest growing model of public cloud computing. Due to the existence of several IaaS cloud providers,

there is increased competition among cloud companies, which develop different pricing models in order to

meet the market demand. As a consequence, prices for cloud services are a result of a multidimensional

function, shaped by the service’s characteristics. The development of a suitable pricing method, based on an

appropriate price index able to capture the market dynamics, is an obvious necessity. The aim of this paper is

the construction of such a price index, for the IaaS model, using data from a wide range of cloud providers

and a large number of price bundles. The hedonic pricing method is used to decompose cloud computing

services into their constituent characteristics, obtaining estimates of the contributory value of each

resource. According to the results, RAM size, CPU power and subscription turned out to be the most

influential factors that shape IaaS pricing.

1 INTRODUCTION

During the recent years, cloud computing has gained

enormous popularity across the business world as

there is an increased demand for a new business

model that can help companies respond faster and

cheaper to their constituents’ needs, not only in

Europe but also worldwide. Its systems and services

are being improved and developed at a fast pace,

offering operational benefits to both the providers and

the consumers of the technology, contributing

substantially at the same time to the creation of a

competitive environment in the global market (Etro

2009).

Therefore, cloud computing is considered to be a

really powerful technological tool and an innovative

business model, composed of three service models:

Infrastructure as a Service (IaaS), which includes

control of fundamental computing resources, such as

memory, computing power and storage capacity;

Platform as a Service (PaaS) that provides control

over the deployed applications and possibly

configuration settings for developer platforms and

Software as a Service (SaaS), which includes the use

of software services accessed through a web browser

or a program interface (Mell and Grance 2011). In

addition to these, cloud computing has also four

deployment models: Private cloud, provisioned for

exclusive use by a single organization; Community

cloud, used exclusively by a specific community of

consumers from organizations that have shared

concerns; Public cloud, open for use by the general

public and Hybrid cloud, which is a composition of

two or more distinct cloud infrastructures (Mell and

Grance 2011).

Concerning the above models, public cloud

computing receives more attention and the IaaS

model gains increased adoption across the business

world (Anderson et al. 2013). More specifically, IaaS,

which is a foundational cloud delivery service and the

most straightforward of the cloud models, provides

flexibility and can be a very good solution for

companies needing computing resources in the form

of virtualized operating systems, workload

management software, hardware, networking, and

storage services (Hurwitz et al. 2012). Computational

power and operating systems are delivered to the

customers in an “on-demand” approach. An

enterprise that migrates its IT system to IaaS may hire

the required resources as needed, instead of buying

499

Mitropoulou P., FIliopoulou E., Tsaroucha S., Michalakelis C. and Nikolaidou M..

A Hedonic Price Index for Cloud Computing Services.

DOI: 10.5220/0005410604990505

In Proceedings of the 5th International Conference on Cloud Computing and Services Science (CLOSER-2015), pages 499-505

ISBN: 978-989-758-104-5

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

them (Mell and Grance 2011). According to Gartner’s

latest report about the public cloud (Anderson et al.

2013), it is expected that IaaS will be the fastest

growing area of public cloud computing achieving a

compound annual growth rate (CAGR) of 41.3%

through 2016, as Figure 1 illustrates.

Figure 1: Public Cloud Services - five year (2011-2016)

CAGRs (%), by model. Source: Gartner (February 2013).

Currently, there are several public cloud providers in

the cloud computing market that provide similar

services to customers. A client's choice of which

cloud company would host his infrastructure-related

services in the long-term depends jointly on the price

it has, the Quality-of-Service (QoS) guarantees it

offers to its customers and the satisfaction of the

advertised guarantees (Vinu Prasad et al. , Siham et

al. 2012). Cloud computing works, in general, on a

“pay-as-you-go” basis, giving the option to the user

to pay for what they use; meaning that the customer

is charged for each computing resource (e.g. RAM,

CPU, storage) separately, usually per unit-hour

(Martens et al. 2012). On the other hand, it is true that

the battle for a dominant market share grows the

competition among cloud companies and leads to the

development of new pricing schemes in order to meet

the market demand. As a result, several packages of

different resources are offered in attractive tariffs and

they are continually fitted to the changing preferences

and increasing needs of customers. Furthermore,

there is an option somewhere in the middle, as there

are some cloud providers who offer predefined sets of

some resources, usually memory and CPU are

bundled, whereas users can select simultaneously on

their own some other computing characteristics, such

as storage size (Martens et al. 2012, Andra 2013).

Finding the right combination of the available

resources is critical for a business to achieve the best

value when creating its own cloud bundle of services.

As any bundle consists of various characteristics,

valued differently by each consumer, there are a

number of questions that arise in the cloud computing

context, such as:

 How customers’ choices and preferences for

IaaS affect the pricing of the corresponding

resources.

 Which characteristics are truly independent

from one another, while at the same time

being the most important for shaping the

pricing of IaaS.

Towards this direction, multiple cloud providers

have already developed cost estimator tools, which

are used to help customers evaluate IaaS services,

deciding the most suitable for their needs. In addition,

another similar approach is considered to be a broker

model that acts as an intermediary between

consumers and providers. Both of these methods are

mainly based on asking users some questions about

the amount of computational power, memory, storage

requirements, data transfer and they subsequently

offer a monthly estimate for price, for the selected

bundle (Hurwitz et al. 2012) or the most cost-efficient

tariff option among many different providers in the

case of a broker (Jörg et al. 2014). However, none of

these tools is capable of identifying the cheapest

cloud hosting provider due to the fact that this is a

choice that depends exclusively on the clients’

computing needs and, furthermore, price is

considered to be a multidimensional function, where

many factors should be taken into account (Siham et

al. 2012).

Into that context, this paper tries to describe the

current mechanism of pricing and reveal the

dynamics which drive the pricing of cloud computing

services. For this, it provides an empirical analysis of

IaaS pricing, by constructing a price index based on a

hedonic pricing method.

The remainder of the paper is structured as

follows: Section 2 presents a brief literature review of

the previous work regarding the pricing methods of

cloud computing, while in Section 3 there is a

theoretical approach of hedonic functions analysis

and price indices. The empirical study and the

evaluation of hedonic regression model together with

discussion are described in Section 4, and finally,

Section 5 concludes, providing directions for future

research.

2 PRICING MODELS OF CLOUD

COMPUTING

In a cloud computing environment, an Infrastructure-

as-a-service demand is considered as an access to the

CLOSER2015-5thInternationalConferenceonCloudComputingandServicesScience

500

system resources, such as CPU, memory and disk.

Resource allocation is a really challenging task

because of the different pricing models the providers

use. In general, cloud hosting providers must offer a

good pricing model that does not bring any loss to

neither the provider nor the consumer and maximizes

the social surplus of the corresponding market. It is

not usually easy to reach a point where both sides

agree with the price set. A user is always willing to

pay a lower price for the resources requested, whereas

a provider should not want to go beyond the lowest

price that gives him no profit (Andra 2013).

Currently, there are many cloud vendors that

follow a fixed pricing strategy, meaning that

customers pay for what they use but not for what the

cloud services value (Vinu Prasad et al.). One of the

most common pricing schemes of this category is the

“pay-as-you-go” model, which charges the users for

just the services they need, paying only for the

required computing instances and just for the time

they use them. In the case they need more resources,

they simply request them from the provider

(Grossman 2009). Another fixed pricing model is

based on subscription, setting a standard price for the

required resources according to a longer period of

subscription (Al-Roomi et al. 2013). However, these

static approaches of pricing have some limitations,

due to the fact that they reserve computational

resources in advance and it is often hard to satisfy

both the cloud vendors’ and cloud users’

requirements.

Hence, dynamic pricing is usually the key solution

to the above problems. It is a method in which the

price for each bundle of resources is based on a

number of factors, such as availability, time, the

service features and according to the forces of the

demand and supply of a real-time market (Andra

2013, Al-Roomi et al. 2013). Mihailescu and Teo

(2010) proposed such an auction-based pricing

strategy for federated clouds, in which resources are

shared among many cloud service providers.

Rohitratana and Altmann (2012) used an agent-based

simulation of four different pricing models that

indicated that the Demand-Driven (DD) pricing

scheme was the best approach in ideal cases. Li et al

(2011) introduced a real-time pricing algorithm for

cloud computing resources. It analyzed some history

utilization data and it found the final price that was

mostly beneficial for the provider because it reduced

its costs, allowing at the same time resources to be

used more effectively. Moreover, there are some

pricing methods that are mostly driven by

competitors’ prices (Rohitratana and Altmann 2010)

and some others based on the amount of money users

are ready to pay (Ruiz-Agundez et al. 2011). All of

the above pricing methods are fair enough for the

customers’ side.

Both of the above categories of pricing models,

especially the dynamic one, take into consideration

some of the service’s most important characteristics.

The construction of price indices is generally used for

this purpose, seeking to estimate the extent to which

each characteristic affects the total price of a service

bundle. Among the most common and widely used

approaches is the hedonic pricing method (Triplett

2004). It was primarily developed seeking to capture

the effect of environmental and housing attributes in

the context of the housing market (Goodman 1978)

and to adjust for quality change for automobiles

(Griliches 1961). In (Chanel et al. 1996) a price index

for paintings, based on regressions using the full set

of sales, was constructed and the idea that goods are

valued for their utility-bearing characteristics can be

found in (Rosen 1974). As far as information and

communication technologies are concerned, hedonic

price indices have been widely used for personal

computers (Pakes 2002, Berndt et al. 1995) taking

quality changes into account and for microcomputers

and printers using evidence from France (Moreau

1996). A hedonic pricing approach has been also

proposed in (Jörg et al. 2014) to estimate price

evolution of telecommunication services based on

data across Europe and in (Siham et al. 2012) the

hedonic pricing method was applied to make cloud

pricing plans more transparent.

3 HEDONIC PRICE INDICES

Hedonic methods refer to regression models in which

a product’s (or a service’s) prices are related to

product characteristics and the observed price of a

product (service) is considered as a function of these

characteristics. The main assumption hedonic

methods are based on, is that a service is a bundle of

characteristics and that consumers just buy bundles of

product characteristics instead of the product itself. A

hedonic method decomposes the item being

researched into its constituent characteristics, and

obtains estimates of the contributory value of each

characteristic, provided that the composite good can

be reduced to its constituent parts and that the market

values those constituent parts.

According to the definition of (Triplett 2004): “A

hedonic price index is any price index that makes use

of a hedonic function. A hedonic function is a relation

between the prices of different varieties of a product,

AHedonicPriceIndexforCloudComputingServices

501

such as the various models of personal computers,

and the quantities of characteristics in them”.

These methods can be used to construct a quality-

adjusted price index of a service. An informative

overview of the hedonic methods and how they are

constructed can be found in (Berndt 1991, Triplett

2004). Moreover, as shown in (Rosen 1974)

consumer chooses from a large number of product

varieties without having the ability to influence

prices. As a consequence, consumers maximize

utility and producers maximize profits. In hedonic

studies it is possible to adjust the price of a service for

its quality not quantity. All of them are based on some

estimated coefficients that are inflicted on the

characteristics of the products in two periods; m and

m + 1. It is possible to estimate the coefficients

separately, for each evaluate period of time, or

consider the observations of two or all periods

together and estimate a common set of coefficients,

seeking to reveal the general trend.

The advantage of this method is that the necessary

calculations are easy to implement. Hedonic methods

are also very fast to apply but the disadvantage is that

index price can change even if no new products exist,

or if all prices remain the same. Among the strengths

of a hedonic pricing method are that it can be used to

estimate values based on actual choices and its

versatility, since it can be adapted to consider several

possible interactions between market goods and

environmental quality.

The hedonic price indices are commonly used as

approximations to find how much money a consumer

would need in period m+1 relatively to the amount of

money required in period m, keeping the same level

of utility. The solution to this problem is to determine

the consumer’s profile and his reaction to a varied and

fast-changing supply of products. The main problem

towards this direction is that each consumer has

potentially different needs and requirements No

matter what profile is decided, it will be a hypothesis

and an assumption that will correspond to a specific

model. In addition to this, a consumer’s desire is not

stable, something quite reasonable since there is a

great offer as technology becomes cheaper and more

attractive.

A hedonic function



, which relates a

number of the product’s characteristics with the

corresponding price as:



fX

(1)

where P

i is the price of a variety (or a model) i of the

considered product and X

i is a vector of

characteristics associated with the specific variety.

The hedonic function is then used, for a number of

different characteristics among the varieties of the

product and the price index is calculated. As soon as

the characteristics to be considered are determined

then, for N varieties of the product (or service) the

following equations must be evaluated:

011 22

··,

1, . . . ,

iiii

bbX bX e



 



(2)

where

are the regression coefficients that have to

be estimated and e

is the regression residual of the

assumed functional form. The regression coefficients

value the characteristics and they are often called

implicit prices, because they indicate the prices

charged and paid for an increment of one unit of the

corresponding characteristic. Implicit prices are much

like other prices, they are influenced by demand and

by supply. In some cases the natural logarithm (ln) of

the price is considered, instead of the actual value.

Furthermore, the functional form of the index can be

nonlinear.

In the case that the prices span between two (or

more) periods of time m and m + 1, the equations to

be evaluated are

011 22

1011 22 1

·· ,

1, . . . ,

·· ,

1, . . . ,

im i i im

PbbXbXe





 



  



(3)

In the context of this work, the vector of

characteristics Xi, corresponds to the configuration of

the IaaS cloud services that affects the price,

including characteristics such as RAM size, number

of CPUs, memory size, bandwidth etc. The

description of these parameters is given in the next

section.

The importance of a price index is that it can be

used to determine suggested prices for combinations

of the characteristics that were not included, or they

were not available, when the index was constructed.

4 PRICE INDEX

CONSTRUCTION

This section describes the empirical study design,

contains the evaluation of the hedonic price index

methodology for cloud computing services, the

construction of a corresponding index and discussion

of the results.

The price index is constructed for the IaaS cloud

CLOSER2015-5thInternationalConferenceonCloudComputingandServicesScience

502

computing, the most straightforward cloud service.

Data collection was based on Cloudorado

(http://www.cloudorado.com), a price comparison

service of cloud computing providers. Cloudorado is

also a price calculator for multiple cloud hosting

providers, since the comparison is performed by

calculating price for individually set server needs. It

currently focuses on providing pricing bundles for

IaaS providers. The number of the collected price

bundles is 2354, by 25 providers, shown in Table 1.

Table 1: Cloud IaaS providers.

Amazon

atlantic.net

Bitrefinery

CloudSigma

Dimensiondata

eApps

ecloud24

Elastichosts

Exoscale

GIGENET

GOGRID

HYVE

JoyentCloud

Lunacloud

Ninefold

OPENHOSTING

Rackspace

SERVERMULE

Storm

StratoGen

Terremark

VPSNET

Windows Azure

Zettagrid

Google is not included among the providers since it

does not offer price bundles but it rather charges for

each CPU and each GB of storage and memory

capacity. The price bundles are specified by the

resources presented in Table 2, together with the

considered values. These characteristics participate as

variables in the hedonic pricing model.

Data correspond to the IaaS services offered by

cloud providers who use different pricing models.

The study started by selecting specific computing

requirements (e.g. 2xCPU, 1GB RAM, 50GB

Storage, 5GB Tranfer-Out, Linux) but due to the fact

Table 2: IaaS characteristics.

Characteristic Description Values

CPU CPU power 2x, 4x, 6x / 3x,

5x, 7x

RAM RAM size in

Gigabytes (GB)

1, 4, 16, 32

Storage Measured in GB 100, 1000

Transfer_Out Number of bytes sent

by server to Internet

per month. (GB)

5, 10000

OS Operating System of

the server

Linux,

Windows

Subscription Indicates if there

should be a

subscription

No, Yes

(corresponds

to 1 year

subscription)

that many of the providers (e.g. Amazon, Rackspace,

GoGrid) use price bundling, the best package of

resources, which was most close to each customer’s

needs, was chosen every time. Prices range between

$31 and $3,318 per month and there are observable

differentiations depending on the existence of a

subscription, while the duration of the subscription

does not affect the price substantially. The operating

system parameter (OS) and the subscription

characteristics participate as dummy variables. The

values for the OS are 0 for Windows and 1 for Linux

and, regarding the subscription, corresponding values

are 0 for no subscription and 1 for a subscription.

Not surprisingly, the most popular geographical

continent for providers is North America, with 19 out

of the 24 to have datacenters located there, followed

by Europe, with 13 providers. Australia and Asia

follow with 8 and 6 providers, respectively, and

Africa comes last with just 1 provider.

Among the limitations of the collected dataset is

that there are a few more characteristics participating

in the construction of the price bundling, which were

not considered into this study. These characteristics

are the Transfer In (the number of bytes received by

server from the internet per month), the Time On

(proportion of the day the server is available) and the

option that the CPUs, the RAM and the storage can

be distributed among more than one physical server.

The value of the Transfer In characteristic does not

contribute at a substantial level to the shaping of the

pricing bundles, because many cloud providers such

as Amazon and ecloud24 charge customers only for

the outgoing traffic and the others include it as a small

amount in the total price of services. Therefore, with

no loss of generality, the Transfer In attribute was

considered to be at 1GB per month. As far as Time

On is concerned it was set at a level of 100%

AHedonicPriceIndexforCloudComputingServices

503

availability per day. The default offered value of non-

distributed resources was also considered.

Moreover, some non-functional factors, such as

availability and reliability of resources or the lack of

fulfilling the agreements between consumers and

providers, were not considered in the price model, as

variables. Inclusion of these characteristics is an

interesting extension of the model, since it may reveal

their potential to affect the price of a cloud offering.

This needs to be empirically proven through the

execution of many different scenarios.

The results of the hedonic pricing method are

summarized in Table 3.

Table 3: Results of hedonic method.

Coefficients Value

Constant

130,499***

(35,42)

CPU

14,532**

(6.41)

Storage

0,249***

(0.02)

RAM

20,434***

(0.86)

-16,91

(2.29)

Transfer_OUT

0,076***

(0.002)

Subscription

-85,82

(20.28)

*** p<.01, **p<.05, *p<.1, n.s. not significant

The calculated R

value is 57.91%, indicating that

although a great portion of the uncertainty is

described by the model, the linear form of the model

may not be the most appropriate to describe the

pricing index and alternative formulations could also

be considered.

As observed, all parameters are significant and

they contribute to the shaping of the price.

Subscription is the parameter contributing more to the

price index, followed by the RAM size and the CPU.

The high value of the constant, which represents a

fixed monthly fee, supports the finding that the

subscription is a crucial parameter. Storage does not

seem to affect the price very much. The choice of the

operating system affects pricing at a high level, since

Linux reduces the price of the bundles by a factor of

16.91.

5 CONCLUSIONS

The hedonic pricing method was used in this work, in

order to develop a price index for the Infrastructure as

a Service cloud computing services. The evaluation

of the method was based on the linear hedonic model

and the data were collected for a number of 22

providers, corresponding to more than 2300 price

bundles.

The results indicate that, apart from the constant

parameter which indicates the importance of the

subscription, a finding that is also supported by the

high value of the subscription parameter, the RAM

size and the CPU are also of substantial importance

and significance. On the contrary, the storage and the

transfer out parameters seem to affect the pricing

procedure less.

As in most cases, there are some certain

limitations in this work, which in turn constitute its

further extension and indicate directions for future

research. Among them is the use of nonlinear

functional forms in the hedonic formulation, seeking

to improve the accuracy of the pricing index. The

value of R

achieved indicate that it would be worth

testing. Apart from the general price index

considering price bundles across all providers, the

construction of an index for each provider would be

of particular interest, mainly for comparison reasons.

The construction of price indices for the other

cloud computing models, namely the software as a

service (SaaS) and the Platform as a Service (PaaS),

where literature has little to present, would be another

important, as well as interesting research direction.

In any case, the existence of a price index for the

cloud services can provide very useful information,

not only regarding the pricing schemes but also

regarding the market of cloud itself and could suggest

optimal pricing approaches of the cloud services.

REFERENCES

Al-Roomi, M., Al-Ebrahim, S., Buqrais, S. and Ahmad, I.

(2013) 'Cloud Computing Pricing Models: A Survey',

International Journal of Grid & Distributed

Computing, 6(5).

Anderson, E., Eschinger, C., Wurster, L., de Silva, F.,

Contu, R., Liu, V., Biscotti, F., Petri, G., Zhang, J. and

Yeates, M. (2013) 'Forecast overview: Public cloud

services, worldwide, 2011-2016, 4Q12 Update',

Gartner Inc., February.

Andra, R. S. (2013) 'Investigating Pricing and Negotiation

Models for Cloud Computing'.

Berndt, E. R. (1991) The practice of econometrics: classic

and contemporary, Addison-Wesley Publishing.

Berndt, E. R., Griliches, Z. and Rappaport, N. J. (1995)

'Econometric estimates of price indexes for personal

computers in the 1990's', Journal of Econometrics,

68(1), 243-268.

Chanel, O., Gérard-Varet, L.-A. and Ginsburgh, V. (1996)

CLOSER2015-5thInternationalConferenceonCloudComputingandServicesScience

504

'The relevance of hedonic price indices', Journal of

Cultural Economics, 20(1), 1-24.

Etro, F. (2009) 'The economic impact of cloud computing

on business creation, employment and output in

Europe', Review of Business and Economics, 54(2),

179-208.

Goodman, A. C. (1978) 'Hedonic prices, price indices and

housing markets', Journal of Urban Economics, 5(4),

471-484.

Griliches, Z. (1961) 'Hedonic price indexes for

automobiles: An econometric of quality change' in The

Price Statistics of the Federal Goverment, NBER, 173-

196.

Grossman, R. L. (2009) 'The case for cloud computing', IT

professional, 11(2), 23-27.

Hurwitz, J., Kaufman, M. and Halper, D. F. (2012) 'Cloud

Services for Dummies', IBM Limited Edition, John

Willy and Sons.

Jörg, G., Hiemer, J. and Hinz, O. (2014) 'A Cloud

Computing Broker Model for IaaS resources', in

Proceedings of the European Conference on

Information Systems (ECIS) 2014, Tel Aviv, Israel,

June 9-11,

Li, H., Liu, J. and Tang, G. (2011) A pricing algorithm for

cloud computing resources, translated by IEEE, 69-73.

Martens, B., Walterbusch, M. and Teuteberg, F. (2012)

Costing of cloud computing services: A total cost of

ownership approach, translated by IEEE, 1563-1572.

Mell, P. and Grance, T. (2011) 'The NIST definition of

cloud computing'.

Mihailescu, M. and Teo, Y. M. (2010) Dynamic resource

pricing on federated clouds, translated by IEEE, 513-

517.

Moreau, A. (1996) 'Methodology of the price index for

microcomputers and printers in France', Industry

Productivity: International Comparison and

Measurement Issues, 99-118.

Pakes, A. (2002) A Reconsideration of Hedonic Price

Indices with an Application to PC's, National Bureau of

Economic Research.

Rohitratana, J. and Altmann, J. (2010) 'Agent-Based

Simulations of the Software Market under Different

Pricing Schemes for Software-as-a-Service and

Perpetual Software', Economics of Grids, Clouds,

Systems, and Services, ser. Lecture Notes in Computer

Science, Altmann et al., Eds. Springer

Berlin/Heidelberg.

Rohitratana, J. and Altmann, J. (2012) 'Impact of pricing

schemes on a market for Software-as-a-Service and

perpetual software', Future Generation Computer

Systems, 28(8), 1328-1339.

Rosen, S. (1974) 'Hedonic Prices and Implicit Markets:

Product Differentiation in Pure Competition', Journal

of Political Economy, 92, 34-55.

Ruiz-Agundez, I., Penya, Y. K. and Bringas, P. G. (2011) A

flexible accounting model for cloud computing,

translated by IEEE, 277-284.

Siham, E. K., Schlereth, C. and Skiera, B. (2012) 'Price

comparison for infrastructure-as-a-service'.

Triplett, J. E. (2004) Handbook on hedonic indexes and

quality adjustments in price indexes, science,

technology and industry working papers, OECD

publishing.

Vinu Prasad, G., Rao, S. and Prasad, A. S. 'A Combinatorial

Auction Mechanism for Multiple Resource

Procurement in Cloud Computing'.

AHedonicPriceIndexforCloudComputingServices

505