A Survey of Cloud Computing Variable Pricing Models

Sahar Arshad

, Saeed Ullah

, Shoab Ahmed Khan

, M. Daud Awan

and M. Sikandar Hayat Khayal

Department of Computer Engineering, College of EME, National University of Sciences & Technology (NUST),

H-12, Islamabad, Pakistan

Faculty of Computer Science, Preston University, Islamabad, Pakistan

Keywords: Cloud Computing, Pricing Models, Service Differentiation, Market-oriented Computing.

Abstract: Cloud computing has grasped the attention of scientific community and business industry towards the

provisioning of computing resources as utility and software as a service over a network. Profitability and

revenue maximization are the most important goals for any cloud service provider which can be employed

through different pricing models. Historically, cloud providers were following the fixed pricing strategies

while provisioning their services to the clients. However, these approaches have their own shortcomings

resulting in resource wastage, lack of fairness and user satisfaction. With expansion of cloud users in market

every day, provisioning of fair resource allocation with service differentiation and efficient pricing model is

demand seeking, resulting in a shift from static to dynamic pricing models. In this paper, we discuss and

investigate different pricing scheme with respect to their advantages, limitations and possible future

directions. The study will open a way for vendors to seek new research directions in dynamic pricing

schemes.

1 INTRODUCTION

Cloud Computing is an emerging IT development,

deployment and delivery model consisting of a

collection of interconnected and virtualized

computers enabling real time delivery of services/

products and solutions over the Internet. It is a

paradigm shift from traditional cluster and grid

computing. Many companies are now taking

advantage by developing their IT transformation

strategy to fasten their time to market by using cloud

pay-per-use services. This new view of building

private network to provide value added services in

economic terms is known as Market Oriented

Computing. Some popular Industrial cloud providers

in market are Amazon EC2 (Elastic Compute

Cloud), Google App Engine, GoGrid, Microsoft

Azure and many more (Shang, 2010).

Cloud computing has drawn increasing attention

in the past few years due to the inherent

characteristics of flexibility, scalability and ease of

use. Just like business organizations, scientific

community is also getting benefits from cloud

capacities by adopting the utility of cloud computing

offered with low cost involved. As more resources

are required for high Performance Computing (HPC)

applications to accomplish their tasks/workload, the

advent of cloud computing has shifted their

computation from the dedicated clusters to the

widely available cloud provided utilities in a pay as

you go fashion (Huang, 2013). It is commonly

perceived by many researchers that in near future, it

will become the fifth utility after electricity, gas,

water and telephone provisioning the people with

computing capacities to use in their daily life

without incurring the capital investments.

Cloud Service Providers (CSPs) offer various

instance types of their Virtual Machine (VM) each

with different price, capacity and computation. CSPs

provide their services to the clients by renting their

spare capacities on hourly, monthly, semi-annual

and annual basis to the clients at comparable offers.

They are liable to manage the peak load capacity and

resource utilization strategically. With the concept of

‘virtualization’ where reallocating resources, to

fulfill the market demands, is managed dynamically

through VM migrations over the network, it has

become easier for providers to manage the load with

negligible cost (Ma, 2012).

Arshad S., Ullah S., Khan S., Awan M. and Hayat Khayal M..

A Survey of Cloud Computing Variable Pricing Models.

DOI: 10.5220/0005429900270032

In Proceedings of the 10th International Conference on Evaluation of Novel Approaches to Software Engineering (ENASE-2015), pages 27-32

ISBN: 978-989-758-100-7

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

Profitability and revenue maximization are the

most important goals for any cloud service provider,

which can be employed through different pricing

models; however, the end-users are typically more

interested in Quality of Service (QoS),

Cost‐effectiveness, Usability and Availability of

cloud resources (Users maximize utility and CSPs

maximize profits). Keeping a balance between these

two (trade-off) is the most challenging design

decision to be made by cloud service providers. In

this research, we are particularly more interested in

pricing policies for cloud computing which may

affect user satisfaction, behavior and loyalty to CSPs

(Al-Roomi, 2013).

Cloud pricing models can be categorised in the

following two general models:

1.1 Pay-as-You Go Pricing/

Pay-per-Use Fixed Pricing

Historically, cloud providers were following the

fixed pricing strategies while provisioning their

services to the clients i.e., clients were charged with

constant unit price for a computing capacity

regardless of the capacity of the compute unit. This

approach has its shortcomings in a way that resource

wastage may occur if the user is not or seldom use a

resource from the set of resources he paid for.

Moreover, vendors are free to charge the user

without taking measures for QoS and other

satisfying metrics, for example, Amazon on-demand

and reserved instances, Google Apps are the

examples of employing such schemes.

The pay as you go billing model by the cloud

providers cut down the ownership cost required to

configure and maintain the real capacities and

storages. Amazon provided the simple price

calculation for a resource as (EC2, 2014):

P =P

comp

storage

out

tran

(1)

where

comp

is the VM instance price. These include

standard, High Memory and High CPU.

storage

is the price charged for storing user data on

cloud.

, P

out

is price associated with uploading and

downloading the data between different regions of

the same cloud.

tran

is the price of file operation within a VM.

This type of pricing scheme is best suited for IaaS.

1.1.1 Pay for Resources

In this technique, user is charged for the use of

storage and bandwidth size accordingly.

1.1.2 Subscription

As the name indicates, user gets subscribe to a

particular CSP with fixed price per unit consumed

for a long period of time (normally 1-3 years) with a

fixed pricing scheme.

Above mentioned pricing models are more

biased towards the providers of the cloud than the

consumers. Thus efficient cloud market mechanism

to deal with maximum number of transactions along

with flexible pricing model to meet the requirements

is the next evolving step that users and providers are

seeking for (Shang, 2010).

1.2 Dynamic Pricing

The environment of the cloud is inherently dynamic.

With the expansion of the cloud users every day in

the market, provision of fair resource allocation with

service differentiation and efficient pricing model is

demand seeking. This has opened a way for vendors

to adopt varying pricing schemes, from static to

dynamic pricing models (Sowmya, 2012). Amazon

was the first to provide the concept of dynamic

pricing model through spot pricing scheme. The

concept was to utilize unused and spare capacity

available in the data centres after fulfilling the

demands of the on-demand and reserved instances.

These unused capacities are referred as spot

instances and are charged based on the fluctuating

supply and demand of these spot instances.

In cloud computing, the provision of leasing

technology as a utility is one of the potential

opportunities to achieve market-based price by the

provider. Thus the concept of auction is the

instrument to achieve this potential benefit from the

market with proper allocation of resources to the

clients. Amazon leads in presenting and adopting the

idea of ‘Auctioning’ the technology as utility in the

market rather than conventional fixed price schemes.

Users can request particular type and number of

instances with willingness of paying maximum price

for it in a particular auction. Cloud providers could

lease their computing capacities in an auction where

in a particular cloud market place, products

advertised in a broker module. Moreover, customers

also submit their demand to this module to fetch the

suitable result for them. In a general auction model,

broker search for the best provider/customer match

with specific parameters and thus provides

opportunity to provide benefit at both ends.

However, there is growing urge of setting the price

offered by providers to bid in a market driven

ENASE2015-10thInternationalConferenceonEvaluationofNovelSoftwareApproachestoSoftwareEngineering

manner; where the prices are set to bid in accordance

with the supply and demand of the instance requests.

Different researchers have opinion that spot prices

set for bidding are not market driven. Rather they

are internally fixed prices; selected randomly from a

small predefined interval. But still auction scheme is

capacities in an efficient manner of allocating

resources with low cost (Lampe, 2012). The current

study is aimed to investigate different research

directions and possible future extensions in dynamic

pricing schemes.

2 RELATED WORK

Cloud computing has grasped the attention of the

industry towards the provision of computing

resources as utility and software as a service over a

network. This is because cloud computing has

combined varying aspects of IT into a business

model that brought revolution in the market. As both

users and providers of cloud are trying to get

acquainted of this paradigm shift, it is important for

them to investigate the pricing models that are

adopted in provisioning the capacities.

a) Salesforce.com for example is leading SaaS

provider and customers are charged monthly

subscription fee for their online CRM software

apps.

b) Amazon EC2, on other hand, charge customers

on hourly basis.

c) CloudSigma, an IaaS provider also charges their

customer on hourly basis with the benefit of

short billing segment.

Amazon collets bid for a spot instance from all the

requesters and decide a spot price. For customers’

convenience, the website maintains the 90 days

history to get idea of the bid prices before they

submit their own. However, the user of the spot

instance has risk of service interruption associated

with him as Amazon periodically updates its bid

price. So the user is safe only of his/her bid price

exceeds the set bid price of the provider. This is not

the case with the on-demand and reserved instances

offered by Amazon. But user has advantage of cheap

price being charged for spot instances to work out

for their low valued jobs. SEOMOZ, a software

vendor used Amazons spot instances for their

services. But due to spot price spike once happened,

Amazon terminated all the services of SEOMOZ.

This accident led SEOMOZ to change their

strategy for job submission strategically and they

used mixed pricing strategy by putting their low

valued jobs/services on spot instances and high

valued jobs/services to on-demand instances. This

infers that service differentiation is important aspect

to be learned by the user in order to adopt the mixed

pricing strategy and get advantage form it (Ma,

2012).

The following table lists some of the studies

based on dynamic pricing along with the research

novelty, methodology, limitations and future

directions that can be further explored.

Figure 1: Price Chart of IaaS Cloud Vendors (Base Plan Price per Hour).

Table 1: Comparison of Cloud Computing Variable Pricing Model.

Paper Title

Objective Methodology Novelty

Tools /software

used

Limitations Future Work

Strategic bidding for

Cloud resources under

Dynamic Pricing

Schemes

(Sowmya, 2012)

To examine the behavior

of the bidders under the

dynamic pricing scheme.

Simulation

Modeling of

strategies for the

bidders in spot

market

Real time spot

data form

Amazon EC2 and

ARENA 12.0

-----

Observing users’

strategic behavior

in dynamic pricing

environment

ASurveyofCloudComputingVariablePricingModels

Table 1: Comparison of Cloud Computing Variable Pricing Model (cont.).

Paper Title

Objective Methodology Novelty

Tools /software

used

Limitations Future Work

CAP

: A Cloud Auto

provisioning

Framework for

Parallel Processing

Using On-demand and

Spot Instances.

(Huang, 2013)

To provide a tool that

helps in reducing the cost

of running the HPC

applications within

deadline by providing the

appropriate cluster size

and cloud instance type

Experimentati

on by

implementing

CAP

on top

of Amazon

EC2

Prediction of

application

performance by

providing the

proper cluster size

to finish job

within its deadline

and cost

Python with

NumPy Lib to

program CAP3

Present

schedule

module of

CAP3 only

supports one

task per

customer

To integrate

sophisticated

bidding strategies

and recovery

mechanism during

pre-emption

Maximizing Cloud

Provider Profit from

Equilibrium Price

Auctions (Lampe,

2012)

To examine the

Equilibrium price auction

allocation problem

(EPAAP) and to provide

heuristic approach in

maximizing the profit of

the cloud provider using

Equilibrium Price

Auctions

Qualitative

Assessment of

the optimal

and heuristic

approaches

Scientifically

addresses the

EPAAP and

presented

heuristic solution

for the

distribution of

VMs

Real time spot

data form

Amazon EC2 for

VM types in

evaluation of

EPAAP

Each bid in

auction is

restricted to

one VM

instance only

To support the

proposed

approaches with

optimized auction

requirements like

live migration of

The pricing model of

cloud computing

services. (Ma, 2012)

To study users’ strategy of

submitting varying jobs

under mixed pricing

scheme for better resource

utilization and benefit for

both user and service

provider.

Numerical

evaluation of

proposed

multi stage

game

Studied the

interruptible

service in a

business context

and using mixed

pricing scheme

with service

differentiation in

cloud computing

----

Job arrival

process of

user in all

stages of the

modeled

game is

identical and

independent

Strategic behavior

of the cloud vendor

in making decision

about pricing

schemes for varying

offered instances.

A knowledge-based

Continuous Double

Auction Model for

Cloud Market

(Shang, 2010)

A model to attain high

market efficiency and

stable trading price in

global cloud resource

market for cloud

interoperability

Simulation -----

Double Auction

Round Algorithm

Only one type

of resource at

a time in a

particular

auction

To apply the

proposed model to

real cloud resource

Environment

Procurement Auctions

to trade computing

capacity in the Cloud

(Di Modica, 2013)

To analyze auction based

marketing mechanism

based on accurate

parameters that can lead to

computing business

opportunities with

increasing revenue.

Simulation of

Auction Cloud

Market

Addressed the

issues related to

the bidder’s

strategy in

procurement

auction context

CloudSim

Simulator

------

To build business

model of

auctioneer(broker)

inorder to

investigate the

profit for all market

actors

Themis: Economy

Based Automatic

Resource Scaling for

Cloud Systems

(Costache, 2012)

A spot market based

system that maximizes the

resource utilization of

dynamic applications

while supporting their

SLO’s

Simulation

A proportional

share auction with

support of SLO

management on

top of this

auction.

CloudSim

Toolkit

-------

To support multiple

resource types in a

single auction

Combinatorial

Auction-Based

Dynamic VM

Provisioning and

Allocation in Clouds

(Zaman, 2011)

To evaluate Combinatorial

Auction against fixed

pricing Cloud strategies

while provisioning

dynamic resource

allocation to solve VM

Allocation problem in

Cloud

Simulation

Experiments

Formulated the

dynamic VM

Provisioning

Problem and

proposed the

combinatorial

auction based

mechanism

CA-PROVISON

Algorithm

evaluated with

real workload

data from Parallel

Workloads

Archive

User can

participate in

an auction

with one job

at a particular

time of

auction

Setting up a private

Cloud with real

implementation of

the CA-

PROVISION

Mechanism

ABACUS: An

Auction-Based

Approach to Cloud

Service Differentiation

(Zhang, 2013)

Automatic service

differentiation of jobs with

different budgets, utility

properties and priorities

while optimally allocating

and scheduling resource

Experimentati

Provided

automatic service

differentiation

along with job

priority of the user

while allocating

resources

Experiments of

ABACUS

components on

Hadoop0.20

Only 5 users

in computing

experiments

on a cluster

To handle

dependent resources

in auction using

dependency model

Cost-Optimal Cloud

Service Placement

under Dynamic

Pricing Schemes 

(Li, 2013)

To investigate cost

optimization problem from

cloud providers

perspective

Experimental

Simulation

-------

Optimus cloud

toolkit

-------

Scheduling

mechanisms with

dynamic pricing

schemes for service

placement in cloud

infrastructure

ENASE2015-10thInternationalConferenceonEvaluationofNovelSoftwareApproachestoSoftwareEngineering

Table 1: Comparison of Cloud Computing Variable Pricing Model (cont.).

Paper Title

Objective Methodology Novelty

Tools /software

used

Limitations Future Work

Provisioning Spot

Market Cloud

Resources To Create

Cost-Effective Virtual

Clusters

(Voorsluys, 2011)

Resource allocation and

job scheduling strategy to

run deadline-constrained

jobs on low cost virtual

clusters

Simulation

Dynamic virtual

cluster by

utilizing spot

instances to run

streamed jobs in a

fast and

economical way

Real price

variation data

form Amazon

Rescheduling

of missing

deadline jobs

is quite cost

effective

To apply fault

tolerance

techniques in case

of interruption from

service of spot

instances.

Optimal Pricing of

Multi-model Hybrid

System for PaaS

Cloud Computing 

(Lu, 2012)

To maximize revenues and

minimize costs by using

an optimal hybrid system

Numerical

Modeling

Demand Curve

and

Tiered pricing

policy to obtain

higher revenue

Mathematical

approach

Job arrival

and exit are

assumed to be

uniform

Job utility and

service

differentiation can

be added to refine

model

Optimal Bidding in

Spot Instance Market 

(Song, 2012)

Optimal bidding strategy

for reduction in

computational cost by

utilizing spot instance

market while maximizing

the profit

Simulation

Bidding algorithm

to leverage spot

instances, while

and maximize

profit

Matlab

Only single

type of spot

instance is

utilized

Dynamically tune

bids for deadline

constrained jobs

On the Cost-QoE

Trade-off for Cloud-

based Video

Streaming under

Amazon EC2’s

Pricing Models 

(He, 2014)

To investigate trade-off

between VM procumbent

cost and achieved quality

of experience of end user

Numerical

Evaluation

Joint problem of

resource provision

and procurement

under multiple

pricing models

Theoretical

Model

Bidding

strategies for

procurement

of Spot VM is

not addressed

------------

3 DISCUSSION & CONCLUSION

Theoretically, fixed pricing scheme is a preferable

cost optimization solution for cloud vendors as

compared to variable pricing scheme (Abhishek,

2012). However, recent research work in this

domain is more directed towards variable pricing

scheme as it involves market dynamics of supply

and demand and run-time cost optimization.

Variable pricing models, as proposed in the

literature, are mostly biased towards cloud providers

by aiming profit maximization through equilibrium

pricing (Shang, 2010; Lampe, 2012; Costache 2012;

Di Modica 2013). However, most research ignored

one important aspect; economic efficiency of cloud

computing resources. Economic efficiency means

that right quantity of products/ services are offered

to the users at minimum cost (Supply and Demand,

2013). Hence, the relationship between highest QoS

and price plays an important role in neoclassical

economics of efficiency. Customers will search for

the CSP with highest level of QoS along with the

minimal cost involved in computation. This is the

reason why Amazon spot instances are generated at

‘random’ within a tight price interval (upper, lower)

instead of the claimed market driven pricing scheme

(Agmon, 2013). We plan to incorporate financial

options theory with cloud pricing schemes to address

this issue and propose a pricing model based on

market dynamics and economic efficiency.

REFERENCES

Abhishek, V., Kash, I. A., & Key, P. (2012). Fixed and

market pricing for cloud services. 7th Workshop on

the Economics of Networks, Systems, and

Computation, pp. 157– 162.

Agmon Ben-Yehuda, O., Ben-Yehuda, M., Schuster, A.,

& Tsafrir, D. (2013). Deconstructing Amazon EC2

spot instance pricing. ACM Transactions on

Economics and Computation, 1(3), pp. 1-20.

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

(2013). Cloud computing pricing models: a survey.

International Journal of Grid and Distributed

Computing, 6(5). pp.93-106.

Amazon Elastic Compute Cloud. (2014), EC2 [Online]

Available from: http://aws.amazon.com/ec2.

[Accessed: 10 March 2014].

Costache, S., Parlavantzas, N., Morin, C., & Kortas, S.

(2012). Themis: Economy-based automatic resource

scaling for cloud systems. 9th International

Conference on High Performance Computing and

Communication (HPCC-ICESS), pp. 367-374. IEEE.

Di Modica, G., Petralia, G., & Tomarchio, O. (2013).

Procurement auctions to trade computing capacity in

the Cloud. (3PGCIC), 8th International Conference on

P2P, Parallel, Grid, Cloud and Internet Computing, pp.

298-305. IEEE.

He, J., Wen, Y., Huang, J., & Wu, D. (2014). On the Cost–

QoE tradeoff for cloud-based video streaming under

ASurveyofCloudComputingVariablePricingModels

Amazon EC2's pricing models. ,IEEE Transactions on

Circuits and Systems for Video Technology, 24(4), pp.

669-680.

Huang, H., Wang, L., Tak, B. C., Wang, L., & Tang, C.

(2013). CAP3: A cloud auto-provisioning framework

for parallel processing using on-demand and spot

instances. Sixth International Conference on Cloud

Computing (CLOUD), pp. 228-235. IEEE.

Lampe, U., Siebenhaar, M., Papageorgiou, A., Schuller,

D., & Steinmetz, R. (2012). Maximizing cloud

provider profit from equilibrium price auctions. 5th

International Conference on Cloud Computing

(CLOUD), pp. 83-90. IEEE.

Li, W., Svärd, P., Tordsson, J., & Elmroth, E. (2013).

Cost-optimal cloud service placement under dynamic

pricing schemes. In Proceedings of the 2013

IEEE/ACM 6th International Conference on Utility

and Cloud Computing, pp. 187-194. IEEE.

Lu, H., Wu, X., Zhang, W., & Liu, J. (2012). Optimal

pricing of multi-model hybrid system for paas cloud

computing. International Conference on Cloud and

Service Computing, pp. 227-231. IEEE.

Ma, D., & Huang, J. (2012). The pricing model of cloud

computing services. 14th Annual International

Conference on Electronic Commerce, 263-269. ACM.

Shang, Shifeng, Jinlei Jiang, Yongwei Wu, Guangwen

Yang, and Weimin Zheng (2010) A knowledge-based

continuous double auction model for cloud market..

Sixth International Conference on Semantics

Knowledge and Grid (SKG), pp. 129-134.IEEE.

Song, Y., Zafer, M., & Lee, K. W. (2012). Optimal

bidding in spot instance market. IEEE Conference on

Computer Communications (INFOCOM), pp.190-198.

Sowmya, K., & Sundarraj, R. P. (2012). Strategic bidding

for cloud resources under dynamic pricing schemes.

International Symposium on Cloud and Services

Computing (ISCOS) pp. 25-30. IEEE.

Supply and Demand: The Market Mechanism (2011)

[Online] Available from:

http://kr.mnsu.edu/~cu7296vs/supdem.htm [Accessed

on March 24, 2014).

Voorsluys, W., Garg, S. K., & Buyya, R. (2011).

Provisioning spot market cloud resources to create

cost-effective virtual clusters. Algorithms and

Architectures for Parallel Processing, pp. 395-408.

Springer Berlin Heidelberg.

Zaman, S., & Grosu, D. (2011). Combinatorial auction-

based dynamic VM provisioning and allocation in

clouds. Third International Conference on Cloud

Computing Technology and Science (CloudCom), pp.

107-114. IEEE.

Zhang, Z., Ma, R. T., Ding, J., & Yang, Y. (2013).

ABACUS: An auction-based approach to cloud service

differentiation. International Conference on Cloud

Engineering (IC2E), pp. 292-301. IEEE.

ENASE2015-10thInternationalConferenceonEvaluationofNovelSoftwareApproachestoSoftwareEngineering