A Research-backed Extended Taxonomy for Cloud Computing Elasticity

Raouia Bouabdallah

Higher Management Institute of Tunis, Tunis University, Bou Choucha Street, Tunisia

Keywords:

Elasticity, Taxonomy, Cloud Computing, Provisioning, Monitoring.

Abstract:

Elasticity is an important feature to characterize cloud computing from traditional Information Technology

(IT) infrastructure. It refers to the ability of the cloud provider to provision and release cloud resources, with

demand, appearing to be inﬁnite in any quantity at any time.In this paper, we propose a taxonomy which is

an extended elasticity classiﬁcations compared to existing ones. Then, we discuss industrial and academic

research on cloud elasticity to identify the main issues and drawbacks. Finally, we propose a synthesis of the

studied works based on elasticity solutions’ characteristics provided from our taxonomy.

1 INTRODUCTION

Cloud computing (Fontana de Nardin et al., 2021) is

a new paradigm in the IT evolution. In the literature,

there is no,universal deﬁnition of the cloud computing

(Chen et al., 2022). It is commonly accepted that the

cloud is characterized by a certain degree of elastic-

ity and on demand access capacity. A deﬁnition given

by the National Institute of Standards and Technolo-

gies (NIST) (Mell and Grance, 2011) deﬁnes cloud

Computing as a model for enabling on-demand net-

work access to a shared pool of conﬁgurable com-

puting resources that can be quickly put into service

through the interaction with the provider. This model

is composed of ﬁve essential characteristics (Karuna

Pande Joshi and Yesha, 2010) which are: on-demand

self-service, broad network access, resource pooling,

measured service (Sambit et al., 2020) and rapid elas-

ticity. On-demand self-service refers to the ability of

a cloud provider to provision cloud resources when-

ever the clients need them. Broad network access de-

scribes the network used to access resources hosted in

the cloud. Resource pooling describes a situation in

which the resources of the provider are pooled among

several clients. Measured service refers to the abil-

ity of the cloud provider to monitor, control and re-

port resources. Elasticity is a unique feature of the

cloud technology(BAR, 2020). It refers to the ability

of cloud provider to provision cloud resources, with

demand, appearing to be inﬁnite in any quantity at

any time (Beltran, 2016). Efﬁcient management is a

challenging task with cloud elasticity. To get a wide

https://orcid.org/0000-0002-6386-4223

view of the main cloud elasticity issues and gaps, a

profound survey of research efforts is required. In this

paper, we will propose an extended elasticity classiﬁ-

cations taxonomy and we will discuss industrial and

academic research on cloud elasticity based on elas-

ticity solutions’ characteristics.

The structure of this paper is as follows. In Sec-

tion 2, we present related studies accordance with the

elasticity. Afterwards, we present, in Section 3, a syn-

thesis of the mentioned works within a classiﬁcation

table. In Section 4, we provide an extended classiﬁ-

cation in accordance with the scope, purpose, policy,

method, monitoring metric, etc.

2 OVERVIEW OF ELASTICITY

SOLUTIONS

To get a broad view of the elasticity problem, a review

of related work is required to identify the main issues

and drawbacks. Doing so, we present in this section

an overview of state-of-the-art solutions. We classify

these solutions regarding the elasticity policies used

for executing of monitoring actions. These policies

are classiﬁed into three models which are: reactive,

proactive and hybrid (e.g. proactive and reactive mod-

els).

2.1 Reactive Elasticity

Many of the research is interested with the reac-

tive elasticity model. For instance, the authors in

Bouabdallah, R.

A Research-backed Extended Taxonomy for Cloud Computing Elasticity.

DOI: 10.5220/0011070500003200

In Proceedings of the 12th International Conference on Cloud Computing and Services Science (CLOSER 2022), pages 231-237

ISBN: 978-989-758-570-8; ISSN: 2184-5042

231

(P. Fawaz and Lionel, 2016) and (Paraiso et al., 2013)

propose a multi-cloud platform as a service (PaaS) ap-

proach based on a reactive model for monitoring the

distributed application through a Monitoring compo-

nent which is deployed on any cloud resource (e,g.

VM) hosting the application. This component is re-

sponsible for capturing any change in the state of the

application as well as collect, aggregate and report in-

formation about an application deployed on multiple

cloud environments. This approach adopts a reactive

model based on a set of elasticity rules to increase or

decrease the number of resources running the appli-

cation via a load balancer used to distribute the ap-

plication’s workloads. However, these rules are not

dedicated to specify the predication expression.

In this work (d. Alfonso et al., 2013), the authors

propose a system called CLUES (Cluster Energy Sav-

ing System) which is a general power management

tool based on Dynamic Power Management (DPM)

approach to optimize the energy consumption of the

cluster . This system uses an automated mechanism

to power on or power off the resources (e.g., vir-

tual machines) according to the current workload and

integrates different resource managers by using Re-

source Manager Connectors. However, these connec-

tors are basically used to the batch systems which pro-

vide control over batch jobs, than the cloud systems

(e.g., opennebula and openstack). The cloud systems

can not provide monitoring information about the de-

ployed resources such as the number of handled jobs

regarding the processing units (cores). The authors,

in this work, propose also a reactive model to han-

dle the energy consumption. This model executes an

action when the measured value is over or under a

given threshold deﬁned by the client. However, the

proposed model does not take into account a time in-

terval that the scale out (down) condition should sat-

isfy before executing an action. This may result in

oscillation of the system.

Another reactive model of elasticity was proposed

by (Mohamed et al., 2016) to optimize the provi-

sion of resources in the cloud. This work proposes

an extension of the Open Cloud Computing Interface

(OCCI) for monitoring and reconﬁguring resources.

To do so, It deﬁnes a list of OCCI Entities and Mix-

ins to enable cloud resources elasticity. This elas-

ticity was deﬁned by adding elasticity rules based

on an event-condition-action approach for monitor-

ing the metric data obtained from cloud infrastructure.

But, this work does not provide a predictive elastic-

ity mechanism to dynamically minimize the resource

consumption over time.

Most of these previous researchers ( such as

(P. Fawaz and Lionel, 2016), (d. Alfonso et al., 2013)

and (Mohamed et al., 2016)) apply threshold-based

rules to perform the reactive model. Most of them

allow clients to set rules for the provisioning of re-

sources based on two threshold values per perfor-

mance metric, which are: the upper threshold (ThrU)

and the lower threshold (ThrL). However, (Koperek

and Funika, 2012) is based on four threshold values

including: the upper threshold (ThrU), slightly below

the upper threshold (ThrbU), lower threshold (ThrL)

and slightly above the lower threshold (ThroL).

RESERVOIR (Resources and Services Virtualiza-

tion without Barriers) is a FP7 project (Rochwerger

et al., 2009) that aims to provide to all clients ser-

vices oriented computing without requiring a large

capital investment in the infrastructure. To do so,

the RESERVOIR refers to the peer-to-peer federated

clouds to distribute data centers owned by separate

providers. When a cloud provider does not have the

needed computational resources to serve its clients, it

rents these ones from another cloud provider. This

project has two main actors which are: the Service

Provider and Infrastructure Provider (Gal

an et al.,

2009). The Service Provider represents the entity that

needs to lease an IT capacity (e.g., hardware) from

a cloud infrastructure provider, instead, it uses an in-

house one. Doing so, it deﬁnes its requirements in a

Service Deﬁnition Manifest (Chapman et al., 2012).

This is used to specify one or more virtual machine

images in a single ﬁle named OVF package which

is based on the DMTF’s Open Virtualization Format

(OVF) standard. The Service Deﬁnition Manifest may

also deﬁne elasticity rules using Key Performance In-

dicators (KPIs). The Infrastructure Provider is the

essence of the RESERVOIR service cloud. It repre-

sents a cloud vendor that provides resources as vir-

tual machines as a service in a pay-as-you-go ﬁnan-

cial model.

2.2 Proactive Elasticity

Some researchers are interested with the proactive

elasticity to adjust the provisioning of cloud re-

sources. The time series analysis approach is a proac-

tive model in nature. This approach is used to predict

a sequence of metric data measured over a time. It

includes a set of forecast methods such as: the mov-

ing average, auto-regression, exponential smoothing

and polynomial regression. For example, the authors

in (Gong et al., 2010) use the moving average as a

forecast method to automatically predict the resources

required to the application. Unfortunately, this work

provides poor results. Others (Roy et al., 2011) use a

second order auto-regressive moving average method

(ARMA) to identify the number of resources used by

CLOSER 2022 - 12th International Conference on Cloud Computing and Services Science

232

the application. It is worth to note that the proactive

model is not only used to identify the workload of the

application as explained previously in (Gong et al.,

2010) and (Roy et al., 2011), but also is used to opti-

mize the number of VM migrations in the cloud data

center. For example, Subhadra and Anil (Shaw and

Singh, 2015) hope, by this migration, to minimize the

number of physical machines with reduce the perfor-

mance degradation comes by unneeded movement of

VMs. The optimization of the number of VM mi-

grations depends on the current as well as the future

CPU utilization of physical hosts using the exponen-

tial smoothing as a forecast method.

2.3 Hybrid Elasticity

A few pieces of research involved both the proac-

tive and reactive models. For instance, the authors in

(Shaw and Singh, 2015) propose reactive and proac-

tive models for resolving bottlenecks of the web ap-

plication in order to satisfy the response time require-

ment. The reactive elasticity method is applied to re-

solve the bottleneck by scaling up the resources re-

quired by the web application. The proactive elastic-

ity method is applied when the allocated resources are

not required during a period of time to scale down the-

ses resources whenever possible. This model is based

on the time series analysis approach and is developed

using the polynomial regression method.

In (Bauer et al., 2019), the authors propose a re-

active and proactive models to scale up and down

cloud resources meeting the SLA. The authors used

the queering theory method as a proactive model to

scale down the resources and the thresholds method

as reactive model to scale them up. This work does

not propose a novel reactive model method. More-

over, the proposed proactive model uses the number

of requests handled per time unit as a control metric

instead of the hardware metrics such as memory and

CPU usage.

In (Urgaonkar et al., 2008), the authors propose

propose a reactive and proactive models. The predic-

tive method is used over long time scales( hours and

days). While, the reactive model is used over short

time scales ( seconds and minutes).

3 SYNTHESIS OF RELATED

WORK

We present an overview of state-of-the-art solutions

((P. Fawaz and Lionel, 2016), (d. Alfonso et al.,

2013), (Rochwerger et al., 2009), AgentRW11 , etc

) that focus on the monitoring. It is worth noting

that these solutions handle partially the problem of

the bottleneck situations, especially the CPU bottle-

neck which is due to the high trafﬁc web application.

Moreover, these situations attempt to provide moni-

toring for applications in cloud. But almost all the

proposed situations give tooling solutions to monitor

cloud applications behavior. It is worth noting that

almost all of these cited works either do not target

monitoring at different models which are reactive and

proactive, or do not provide an efﬁcient solution to do

that. In Figure 1, we propose a synthesis of the stud-

ied works based on elasticity solutions’ characteris-

tics provided from our taxonomy that we will detail

in the next section.

4 ELASTICITY TAXONOMY

Through the analysis of the studied solutions, we

note that these solutions follow different methods,

strategies and techniques so as to build their mech-

anisms. A review of related studies ((Coutinho et al.,

2015), (Najjar et al., 2014), (BAR, 2020)) in neces-

sary to identify the different classiﬁcations that have

been suggested in accordance with the elasticity so-

lutions’ characteristics. We push further these studies

((Coutinho et al., 2015), (Najjar et al., 2014)) and we

propose a taxonomy which is an extended elasticity

classiﬁcations compared to existing ones. This taxon-

omy is summarized in Figure 2 and is based on the

following characteristics: the scope, purpose, policy,

method, provider, monitoring metric and standard.

4.1 Scope

This characteristic deﬁnes in which cloud layers the

elasticity actions are applied (Al-Dhuraibi et al.,

2018). The decisions made by the cloud provider

or by the user of the cloud technology can be exe-

cuted either in the infrastructure or the application-

platform levels. We assume that these decisions are

represented by the actions of the provisioning of new

resources and releasing the unused ones. When these

decisions are performed at the infrastructure level, the

provider, which has an elasticity controller, converts

the client’s requirement to actions based on the virtu-

alization technology like VMs or containers. In case

the decisions of elasticity actions are performed on

the application or platform level, the elasticity con-

troller is embedded in the application which can be

either one tier or multi-tiered (Herbst et al., 2013).

This controller interacts with the cloud infrastructure

so as to add or release resources.

A Research-backed Extended Taxonomy for Cloud Computing Elasticity

233

4.2 Purpose

The elasticity solutions have different purposes which

are: increase (or decrease) the infrastructure capacity,

energy, performance and cost. Increase or decrease

the infrastructure capacity refers to the strategy ap-

plied by the cloud provider to re-size its infrastructure

capacity according to the amount of clients using this

one. The energy purpose aims at optimizing the total

energy consumption of IT equipments by reducing

the amount of servers used to run virtual machines.

The cost purpose is closely related to the energy

one but also covers the operational cost of such a

resource, hosted in the cloud, with maintaining the

required performance. The performance refers to

the ability of the cloud provider to retain Quality

of Service (QoS) metrics so as to respond to the

client’s requirement. By getting a broader view to the

elasticity objectives, we notice that there are a plenty

of perspectives. The cloud IaaS providers look to

maximize their proﬁt in accordance with minimizing

its infrastructure. The cloud PaaS providers search

to reduce their payments. The clients (end-users)

seek to minimize the cost they pay to cloud provider,

while still receiving a good Quality of Service (QoS).

4.3 Policy

The policy refers to the required interactions for the

execution of elasticity actions. Regarding related

studies (Pham, 2016), there are two types of poli-

cies : the manual and automatic. The manual pol-

icy is performed by a manual intervention from the

client which is responsible for provisioning and mon-

itoring his virtual environment. The provision of a

service in the cloud has ﬁve phases executed in a se-

quential order (Karuna Pande Joshi and Yesha, 2010),

which are: the requirements, discovery, negotiation,

composition and consumption. In the requirement

phase, the client details his requirements. Then, the

provider searches the cloud services corresponding to

the client’s need in the discovery phase. Afterwards,

the negotiation phase occurs when there is conﬂicts

between the cloud provider and client about a service.

In the composition phase, the provider combines the

services required by the client to provide a single ser-

vice. Finally, this service is delivered to the client

in the consumer phase. It is worth to note that the

manual policy is used in several cloud providers such

as Rackspace and Microsoft Azure. However with

the automatic policy, all the elasticity actions are per-

formed automatically, which could be classiﬁed into

the reactive and proactive models.The reactive model

Studied Solutions

Scope Purpose Policy Method Provider Monitoring Metric Standard

Infrastructure

Applicaton

Dec/Infrastructure

capacity

Energy

Performance

Manual provisioning

Automatic

Coarse-grained

Fine-grained

Single

Federated

Application

OCCI

OVF

Requirements

Discovery

Negotiation

Composition

consumption

Proactive

Reactive

Horizontal

Vertical

Centralized

Peer-to-peer

SoCloud

(P. Fawaz and Lionel, 2016)

(Paraiso et al., 2013) X X X X X X X X X

CLUES

(d. Alfonso et al., 2013) X X X X X X

(Mohamed et al., 2016) X X X X X X X X

(Shaw and Singh, 2015) X X X X X X X

(Bauer et al., 2019) X X X X X X X

RESERVOIR

(Rochwerger et al., 2009) X X X X X X X X

Figure 1: Synthesis of related work.

CLOSER 2022 - 12th International Conference on Cloud Computing and Services Science

234

reacts based on thresholds or rules, due to the ob-

served workload changes. Several industrial cloud

providers use this model such as: Amazon EC2 and

Rightscale, as well as academic researches such as:

(Han et al., 2014) and (Liu et al., 2013). Whereas,

the proactive model reacts to the predicted workload

changes. Most of the reviewed works ( such as (Da-

woud et al., 2011), (Liu et al., 2013) and (Sharma

et al., 2011)) that used this model could ﬁt in one of

these three groups of technique: the time series anal-

ysis, queuing theory and control theory.

4.4 Method

It is related to the applied methods to increase (or

decrease) the size of cloud resource. With regard to

the underlining studies (Pham, 2016), there are two

kinds of methods : the coarse-grained or ﬁne-grained.

The coarse-grained scaling refers to the ability of the

cloud provider to provision resources from an external

one. By the way, the ﬁne-grained refers to the ability

of the cloud provider to scale resources in the fron-

tier of its infrastructure capacity. Generally within the

same provider, the resources can be scaled in two dif-

ferent manners: vertical and horizontal. The vertical

elasticity consists in increasing or decreasing charac-

teristics of resources (e.g virtual machine instances)

such as the CPU, memory, etc. However, the hori-

zontal elasticity refers to the ability to increase or de-

crease the number of allocated resources (e.g virtual

machine instances) needed to run an application in the

cloud provider.

4.5 Provider

It refers to the number of cloud providers that the elas-

tic solutions support simultaneously, which could be

a single cloud or a federation of clouds. A single

provider means that the elasticity control is applied

on only one cloud provider. While in case of feder-

ated clouds, the elasticity control is executed across

a set of cloud providers. The examination of review

works show us that there are two cloud federation

models: vertical federation and horizontal federation

(Celesti et al., 2010). The vertical federation reaches

all levels of cloud service models such as IaaS, PaaS

and SaaS. For instance, a SaaS cloud provider puts

his services on the top of a PaaS cloud provider like

Microsoft Azure Service Platform and Google App

Engine. However, the horizontal federation of clouds

deals with one cloud service model level such as IaaS.

4.6 Monitoring Metric

The monitoring metrics describe the types of values

the elasticity control use to monitor the behavior of

the underling cloud resources. The monitoring met-

rics are available at these levels: the operating system

and application metrics. The operating system met-

rics are applied in evaluating the performance of a vir-

tual hardware with overlooking the applications run-

ning on it. These metrics are: the CPU, RAM, Disk

space and network. However, the application metrics

refer to the units of work that involve on the virtual

hardware as well as applications. These metrics can

be the response time and the number of requests.

4.7 Standard

It refers the standards that the elastic solutions sup-

port, which can be the OCCI

and OVF

. The Open

Cloud Computing Interface (OCCI) is one of the ﬁrst

open extensible standards for managing any kind of

resources provided by cloud providers. This standard

is supported by a large community of cloud providers

including commercial ones such as Amazon EC2, as

well as cloud platforms such as CloudStack, Euca-

lyptus, OpenNebula. While, the Open Virtualization

Format (OVF) (DMTF, 2015) speciﬁcation offers a

portable and an extensible format for packaging and

distributing cloud resources (e.g virtual machines) in

a standard format.

5 CONCLUSIONS

In this paper, we discussed industrial and academic

researches on the elasticity in the cloud computing.

Then, we presented a synthesis of the mentioned

works within a classiﬁcation table. Afterword, we

proposed an extended elasticity classiﬁcations taxon-

omy which is an extended elasticity classiﬁcations

compared to existing ones. This taxonomy is based

on the following characteristics: the scope, purpose,

policy, method, provider, monitoring metric and stan-

dard.

In our future work, we will limit our research to

the coarse grained elasticity and we will propose an

extension for the Open Cloud Computing Interface

(OCCI) to support the automatic negotiation between

different cloud providers.

OCCI: Open Cloud Computing Interface

OVF: Open Virtualization Format

A Research-backed Extended Taxonomy for Cloud Computing Elasticity

235

OCCI

OVF

Standard

CPU

RAM

Both

Method

Fine-grained

Coarse-grained

Horizontal

Ver!cal

Memory

CPU

Disk

Network

Nb. of requests

Response !me

Applica!on

Monitoring

Metric

Elas!city

Scope

Infrastructure

Applica!on/ Pla"orm

Increase/Decrease

infrastructure capacity

Energy

Performance

Cost

Purpose

Over-provisioning

Under-provisioning

Proac!ve

Reac!ve

Threshold-based rules

Time series analysis

Queuing theory

Control theory

Fixed gain controllers

Adaptive controllers

Kalman filter

Fuzzy model

Moving average methods

Exponential smoothing

Regression

ARMA

Discovery

Nego!a!on

Composi!on

Requirements

consump!on

Policy

Manual provisioning

Automa!c

Centralized

Peer-to-peer

Single

Federated

Provider

Ver!cal federated

Horizontal federated

Figure 2: Classiﬁcation of elasticity solutions.

ACKNOWLEDGEMENTS

This research was enabled in part by support provided

by SSOIE-COSMOS laboratory from university of

Tunis (Tunisia). We would also like to thank Dso Ser-

vices company by providing us the required resources

to release this work.

REFERENCES

(2020). The views, measurements and challenges of elas-

ticity in the cloud: A review. Computer Communica-

tions, 154:111–117.

Al-Dhuraibi, Y., Paraiso, F., Djarallah, N., and Merle, P.

(2018). Elasticity in cloud computing: State of the art

and research challenges. 11(2):430–447.

CLOSER 2022 - 12th International Conference on Cloud Computing and Services Science

236

Bauer, A., Herbst, N., Spinner, S., Ali-Eldin, A., and

Kounev, S. (2019). Chameleon: A hybrid, proactive

auto-scaling mechanism on a level-playing ﬁeld. vol-

ume 30, pages 800–813.

Beltran, M. (2016). Becloud: A new approach to analyse

elasticity enablers of cloud services. Future Genera-

tion Computer Systems, 64:39–49.

Celesti, A., Tusa, F., Villari, M., and Puliaﬁto, A. (2010).

How to enhance cloud architectures to enable cross-

federation. In 2010 IEEE 3rd International Confer-

ence on Cloud Computing, pages 337–345.

Chapman, C., Emmerich, W., M

arquez, F. G., Clayman, S.,

and Galis, A. (2012). Software architecture deﬁnition

for on-demand cloud provisioning. In Cluster Com-

puting, volume 15, pages 79–100.

Chen, Y., Huo, J., Li, X., Bi, K., Ma, N., Jing, Y., and

Ma, X. (2022). Classiﬁcation and characteristic anal-

ysis of the clouds and dust in a dust-carrying pre-

cipitation process based on multi-source remote sens-

ing observations. Atmospheric Pollution Research,

13(1):101267.

Coutinho, E. F., de Carvalho-Sousa, F. R., Rego, P. A. L.,

Gomes, D. G., and de Souza, J. N. (2015). Elasticity

in cloud computing: a survey. In Annals of Telecom-

munications, page 1–21.

d. Alfonso, C., Caballer, M., Alvarruiz, F., and Hernandez,

V. (2013). An energy management system for cluster

infrastructures. In Computers & Electrical Engineer-

ing, volume 39, pages 2579 – 2590.

Dawoud, W., Takouna, I., and Meinel, C. (2011). Elas-

tic vm for cloud resources provisioning optimizatio.

In In Advances in Computing and Communications.

Springer, page 431–445.

DMTF (2015). Open virtualization format speciﬁcation.

Fontana de Nardin, I., da Rosa Righi, R., Lima Lopes, T. R.,

Andr

e da Costa, C., Yeom, H. Y., and K

ostler, H.

(2021). On revisiting energy and performance in mi-

croservices applications: A cloud elasticity-driven ap-

proach. Parallel Computing, 108:102858.

Gal

an, F., Sampaio, A., Rodero-Merino, L., Gil, V., and Va-

quero, L. M. (2009). Service speciﬁcation in cloud en-

vironments based on extensions to open standards. In

Proceedings of the Fourth International ICST Confer-

ence on COMmunication System softWAre and mid-

dlewaRE, COMSWARE ’09, pages 19:1–19:12, New

York, NY, USA. ACM.

Gong, Z., Gu, X., and Wilkes, J. (2010). Press: Predictive

elastic resource scaling for cloud systems. In 2010 In-

ternational Conference on Network and Service Man-

agement, pages 9–16.

Han, R., Ghanem, M. M., Guo, L., Guo, Y., and Osmond,

M. (2014). Enabling cost-aware and adaptive elastic-

ity of multi-tier cloud applications. In Future Gener-

ation Computer Systems, volume 32, page 82–98.

Herbst, N., Kounev, S., and Reussner, R. (2013). Elasticity

in cloud computing: What it is, and what it is not.

pages 23–27.

Karuna Pande Joshi, T. F. and Yesha, Y. (2010). Integrated

lifecycle of it services in a cloud environment (de-

tailed paper). In In proceedings of The Third Interna-

tional Conference on the Virtual Computing Initiative

(ICVCI 2009), Research Triangle Park, NC. IBM.

Koperek, P. and Funika, W. (2012). Dynamic business

metrics-driven resource provisioning in cloud envi-

ronments. In Parallel Processing and Applied Math-

ematics, pages 171–180, Berlin, Heidelberg. Springer

Berlin Heidelberg.

Liu, Z., Wang, S., Sun, Q., Zou, H., and Yang, F. (2013).

Cost-aware cloud service request scheduling for saas

providers. In The Computer Journal, page 291–301.

Mell, P. and Grance, T. (2011). The NIST Deﬁnition of

Cloud Computing. NIST, national institute of stan-

dards and technology special publication 800-145 edi-

tion.

Mohamed, M., Belaid, D., and Tata, S. (2016). Extending

occi for autonomic management in the cloud. In Jour-

nal of Systems and Software, volume 122, pages 416

– 429.

Najjar, A., Serpaggi, X., Gravier, C., and Boissier, O.

(2014). Survey of elasticity management solutions in

cloud computing. In in Continued Rise of the Cloud,

page 235–263.

P. Fawaz, M. P. and Lionel, S. (2016). socloud: A service-

oriented component-based paas for managing porta-

bility, provisioning, elasticity, and high availability

across multiple clouds. In Computing, volume 98,

pages 539–565.

Paraiso, F., Merle, P., and Seinturier, L. (2013). Managing

elasticity across multiple cloud providers. In Proceed-

ings of the 2013 International Workshop on Multi-

cloud Applications and Federated Clouds, pages 53–

60, New York, NY, USA. ACM.

Pham, M. L. (2016). Roboconf : an Autonomic Plat-

form Supporting Multi-level Fine-grained Elasticity

of Complex Applications on the Cloud.

Rochwerger, B., Breitgand, D., Levy, E., Galis, A., Nagin,

K., Llorente, I., Montero, R., Wolfsthal, Y., Elmroth,

E., Caceres, J., Ben-Yehuda, M., Emmerich, W., and

Galan, F. (2009). The reservoir model and architec-

ture for open federated cloud computing. In IBM Jour-

nal of Research and Development, volume 53(4), page

4–1.

Roy, N., Dubey, A., and Gokhale, A. (2011). Efﬁcient

autoscaling in the cloud using predictive models for

workload forecasting. In 2011 IEEE 4th International

Conference on Cloud Computing, pages 500–507.

Sambit, K.-M., Bibhudatta, S.-P., and Paramita, P.

(2020). Load balancing in cloud computing: A

big picture. Journal of King Saud University -

Computer and Information Sciences, 32(149-158).

https://doi.org/10.1016/j.procs.2015.03.168.

Sharma, U., Shenoy, P., Sahu, S., and Shaikh, A. (2011).

A cost-aware elasticity provisioning system for the

cloud. In in 2011 31st International Conference

on Distributed Computing Systems (ICDCS), page

559–570.

Shaw, S. B. and Singh, A. K. (2015). Use of proactive

and reactive hotspot detection technique to reduce the

number of virtual machine migration and energy con-

sumption in cloud data center. In Computers & Elec-

trical Engineering, volume 47, pages 241 – 254.

Urgaonkar, B., Shenoy, P., Chandra, A., Goyal, P., and

Wood, T. (2008). Agile dynamic provisioning of

multi-tier internet applications. In ACM Trans. Auton.

Adapt. Syst., volume 3, pages 1:1–1:39, New York,

NY, USA. ACM.

A Research-backed Extended Taxonomy for Cloud Computing Elasticity

237