Emerging Technologies and Innovation: Auto-Scaling in Cloud Computing

This report delves into the auto-scaling technique, a crucial aspect of cloud computing that provides elasticity to applications. It begins with an introduction to the increasing popularity of cloud computing and its elastic nature, which allows users to dynamically acquire and release resources. The report then provides a comprehensive literature review, exploring various studies on auto-scaling, including challenges related to monitoring, analysis, and execution. It examines system components like virtual cluster monitors, front-end load balancers, and auto-provisioning systems. The report also includes a classification of auto-scaling features and identifies key criteria. Finally, it discusses system validation using techniques like threshold-based policies, queuing theory, and control analysis. The report concludes by highlighting the importance of auto-scaling for efficient resource allocation and cost optimization in cloud environments. This analysis is supported by findings from various research papers, emphasizing the significance of auto-scaling in meeting dynamic workload demands and mitigating price competition.

Running head: EMERGING TECHNOLOGIES AND INNOVATION
Emerging Technologies and Innovation
Name of the Student
Name of the University
Author Note

1EMERGING TECHNOLOGIES AND INNOVATION
Abstract:
The main aim of this paper is discussing the auto-scaling technique for providing the elasticity
feature to the applications running in the cloud computing environment. Here, a deep insight has
been gained of the auto-scaling feature from the literature survey done in this paper. The auto-
scaling feature comprises some important system components including the virtual cluster
monitor system, front-end load balancer and auto-provisioning system. Classification regarding
this auto-scaling feature of cloud computing has been also done and from there five important
criteria has been identified. Proper validation of the overall system for the auto-scaling feature
has been also done in this report.

2EMERGING TECHNOLOGIES AND INNOVATION
Table of Contents
1. Introduction:................................................................................................................................3
2. Literature Review:.......................................................................................................................4
3. System Components:...................................................................................................................9
3.1 Virtual Cluster Monitor System:.........................................................................................11
3.2 Front-End Load Balancer:...................................................................................................12
3.3 Auto-Provisioning System:..................................................................................................12
4. System Classification:...............................................................................................................13
5. System Verification:..................................................................................................................14
5.1 Threshold-Based Policies:...................................................................................................15
5.2 Queuing Theory:..................................................................................................................15
5.3 Reinforcement Learning:.....................................................................................................15
5.4 Time-Series Analysis:..........................................................................................................16
5.5 Control Analysis:.................................................................................................................16
6. Discussion:.................................................................................................................................17
7. Conclusion:................................................................................................................................18
8. References:................................................................................................................................19

3EMERGING TECHNOLOGIES AND INNOVATION
1. Introduction:
In present aspect cloud computing is one of the emerging technologies which is
becoming more popular day by day. One important reason behind this much popularity of the
cloud computing is the elastic nature of it [1]. This elastic nature of cloud computing provides a
freedom to the users that they easily acquire and release resource whenever they require. In this
way they only need to pay for the required resources. Many organizations use the application of
cloud computing for different of purpose where the most common application is the storage and
backup purpose.
It is already defined that one of the important characteristics of cloud computing is
elasticity. Elasticity helps the user to release and acquire resources in a dynamic way so that if
there is any changes in demand it can be managed very easily [2]. Though elasticity of cloud
computing is very much beneficial, still deciding right amount of resource is quite difficult task.
Proper allocation of the resources is critical issue regarding cloud computing. For the predictable
situations resources can be provided in advance but for the unplanned situations an automatic
scaling system or an auto-scaling system will be required which is capable of adjusting the
allocated resources to an application at any instance of time. In this way, resources can be
allocated automatically as per the demand of the resources.
In this report a brief research will be done regarding this auto-scaling technique for the
elastic application in cloud environment. For performing this research, a vast literature survey
will be done in this paper and the classification of the system will be done. Important finding
from the research will be also presented in this report.

4EMERGING TECHNOLOGIES AND INNOVATION
2. Literature Review:
Currently, several of previous studies are available which has focused on the elasticity
nature of the cloud computing. In this section a brief literature survey will be done regarding the
auto-scaling technique within cloud environment.
According to the author Qu, Calheiros and Buyya (2018), web application provides are
currently migrating their applications direct to the cloud data centres due to the advanced
facilities provided by the cloud computing paradigm. One of these advanced facilities is the
elasticity of the cloud computing. By using the elasticity feature, the users are able to release or
acquire cloud computing resources on their demand [3]. In this way the web application
providers can automatically scale the provided resources to their applications without any type of
human intervention at the point of some dynamic workload. In this overall cost of the resources
can be minimised while quality of service requirements can be satisfied easily. Here, the authors
have actually analysed the main challenges which are associated with the auto-scaling web
applications regarding the cloud computing and the authors have also reviewed the developments
that are already done within this field. The authors have presented a taxonomy of auto-scaling as
per the key properties and challenges that has been identified. Also, surveyed works has been
analysed by the authors and it is mapped with the taxonomy for the identification of weaknesses
in this field. Depending on the analysis the authors have also proposed future directions for
exploring this area. In this aspect various of challenges is faced by the auto-scaling technique.
These challenges are mainly associated with the monitoring, analysis, planning and execution.
The taxonomy that has been developed by the authors is capable of covering application
architecture, adaptivity, session stickiness, resource estimation, scaling indicators, scaling
timing, oscillation mitigation, environment and scaling methods.

5EMERGING TECHNOLOGIES AND INNOVATION
As per the authors Guo, Stolyar and Walid (2018), the problem regarding autoscaling for
application hosting within a cloud environment can be considered. In this aspect the application
are elastic and is capable of changing the request number with the time. The requests made by
the applications are served by the virtual machines and these resides within physical machine
within a cloud environment. In this aspect the aim of the authors is minimising the overall
number of physical machine hosting and it will be done by intelligently packing the virtual
machines within the physical machines while the virtual machines will be auto-scaled. Regarding
this problem the authors has used a shadow routing-based approach [4]. A special shadow
routing algorithm has been proposed by the authors which is capable of employing special
constructed virtual queuing system so that an optical solution can be produced dynamically. This
will be guiding both of the virtual machine autoscaling and virtual machine to physical machine
packing. The algorithm proposed by the authors is capable of running continuously without any
need of resolving the problem associated with the optimization from scratch. This algorithm also
adapts by itself if there is any type of changes regarding the demand of the application. A
simulation experiment has been also done by the authors. From the simulation experiment it has
been assessed that the algorithm is performing good and the adaptability of the algorithm is also
high.
The authors Rodriguez and Buyya (2018), containers are considered as self-contained and
standalone units which packages both the software and the corresponding dependencies together.
Here, it offers a lightweight performance isolation, flexible and fast implementation and
appropriate resource sharing. Containers have gained high amount of popularity due to better
application deployment and management. This has become popular as it is very much useful for
the diverse workloads including the web services. Due to this, separate, container orchestration

6EMERGING TECHNOLOGIES AND INNOVATION
platform has been developed. These platforms are capable of optimising the schedulers of the
containers on a private cluster of fixed size [5]. But these are not able to auto scale the cluster
size for considering specific features to the public cloud environment. Here, the authors have
proposed an approach regarding comprehensive resource management. This approach is
consisting three different types of objectives. The first objective is optimising initial phase of the
containers by appropriate scheduling them with the existing resources. The second objective is
autoscaling overall resource number during the runtime depending on the cluster workload. The
third objective is rescheduling the mechanism which will be helping in efficient utilization of the
resources.
According to the authors Singh et al. (2019), cloud computing provides an emerging
environment which attracts various of application providers for the deployment of web
applications on the cloud data centres. The main reason due to which the application providers
are very much attracted to the cloud computing is the elasticity provided by the cloud computing
environment. The elasticity is very much useful as is provides the auto-scaling facility which
provides resources on demand [6]. Most of the web applications consists of dynamic workload
and due to this it becomes hard to predict total amount of required resources. In the current
aspects the cloud service providers are also working to reduce the overall cost of using the cloud
computing services without compromising with the service quality. Regarding this one of the
important issues is the auto-scaling. The auto-scaling feature of cloud computing is still in the
infancy and it requires some detailed investigation of taxonomy. Here, the authors have actually
presented a detailed literature survey regarding the auto-scaling techniques of web-based
applications within the cloud computing environment. Through this survey the researchers have

7EMERGING TECHNOLOGIES AND INNOVATION
supported the research community for the identification of auto-scaling technique requirements.
The authors have also proposed some new research areas within this direction.
As per the authors Fazli, Sayedi and Shulman (2018), web-based firms are quite relied
over the cloud-based computational resources so that they can provide appropriate service to the
customers. Here, these organization might be aware about what type of cloud-based
computational resources they will be required but they are very much unaware about total
amount of resources they will be required at instance of time. This is one crucial issue that is
related with the cloud computing environment. To solve this issue, a recent innovation has been
done within cloud computing aspects and it is known as the auto-scaling feature. The auto-
scaling feature provide the functionality of automatic scaling up the available computational
resources so that resources required for any type of processes in the cloud can be matched easily
[7]. In this way proper computational resources can be allocated whenever required. The authors
have actually developed a game theory model for examining how the auto-scaling will be able to
decision of a firm regarding entering into a new market that is based on the cloud computing
services. This model produced by the authors is capable of producing novel results dependent on
the success of a firm within the new market. From this model the authors have successfully
shown that autoscaling is capable of mitigating price competition if the cost of the capacity is
low in a sufficient manner. This happens due to the fact that without the auto-scaling feature
organizations become very much uncertain about the resource demand and it leads to excessive
usage of the computational resources and that ultimately creates an aggressive price competition.
The authors have also found that success rate of newly entered organization in the cloud
computing market becomes very much high by the utilization of auto-scaling feature.

8EMERGING TECHNOLOGIES AND INNOVATION
The authors Aldossary and Djemame (2018), have elaborated that virtual machine auto-
scaling is quite an important technique for the provision of additional resource capacity within a
cloud computing environment. The auto-scaling allow the virtual machine to dynamically
decrease or increase the required amount of resources so that the quality of the service can be
maintained. However, the authors have founded that this auto-scaling feature can be time
consuming while imitating and that is totally unacceptable for the virtual machines which need to
be scale up for any kind of computational processes [8]. Also, there is some additional cost
associated with it which occurs due to increased energy utilization for those extra resources.
Regarding this, in this paper the authors have introduced energy and performance-based cost
prediction framework so that total cost of virtual machines auto-scaling can be estimated by only
consideration of power usage and resource consumption. Here, any type of compromise will be
not done with the performance. Here, the researcher has also assessed that the developed
framework is capable of auto-scaling workload prediction while the cost-saving can be done up
to 25%.
As per the authors Podolskiy, Jindal and Gerndt (2018), one of the important features of a
cloud-based application is the scalability of it. Major cloud service providers who provide IaaS
service uses the auto-scaling feature on the virtual machine level. The other solutions related
with the virtualization can be scaled also. A specific type of application can be scaled depending
on the response to changes within the observed metrics which can include utilization of network,
CPU, storage, etc. Though the auto-scaling feature provides automatic utilization of the
resources during any type of computation it can still fail to meet the quality of service [9]. This
happens in some rare cases due to scaling which occurs by the reactivity of the auto-scaling
solutions. In this paper, the authors have evaluated the performance of auto-scaling for two

9EMERGING TECHNOLOGIES AND INNOVATION
layered virtualizations. This evaluation has been done using the AWS, Google and Microsoft
which are the publicly available clouds. For this evaluation, autoscaling performance
measurement tool has been used by the authors which is also developed by the authors.
According to the authors Nadjaran Toosi et al. (2019), it is very much anticipated that the
future network will be having support for important network functions which includes load
balancers, firewalls and the intrusion prevention systems where all of these will be working in a
flexible, effective and a full automated manner. Within the cloud computing environment, the
network function virtualization is aiming to reduce the costs while simplifying the network
related operations using the virtualization technology. For enforcing the network policies within
the network function virtualization based cloud environment, network services composing virtual
network functions are chained with each other as the service function chain [10]. A policy that
has been matched by the network traffics must be traversing network functions in a chain
sequence for complying with it. Here, the service function chains has drawn a high amount of
attention as compared with provided to dynamic auto-scaling of the virtualized network
functions resources among the service chain. The authors, have assessed that most of the already
existing approach is only focusing on the allocation of the network and the computing resources
to the virtual function network without any type of consideration of service quality of the service
chain including end-to-end latency. From all the aspects the authors have defined unified
framework so that elastic service chain can be established. Also, the authors have introduced a
dynamic algorithm of auto-scaling that is known as ElasticSFC. As per the authors through this
algorithm cost can be minimised while end-to-end latency can be achieved of the service chain.
From the experimental results it has been assessed that proposed algorithm by the authors is
capable of reducing the overall cost of service function chain deployment.

10EMERGING TECHNOLOGIES AND INNOVATION
3. System Components:
Considering any type of web service within a cloud computing environment, it should be
available at instance of time and it should be providing fast response to each of the users
regardless the number of users connected to the cloud computing environment [11]. For this
reason, the system needs to be scalable where it will be able to allocate the computing resources
dynamically as per the requirements of the web applications. Considering this, auto-scaling was
introduced for solving this problem. Here, the auto-scaling system consists three main
components. These components are the virtual cluster monitor system, front-end load balancer
and auto-provisioning system consisting an auto-scaling algorithm [12]. In the following section
all of these components are evaluated briefly.
(Figure 1: System Demonstration of Auto-Scaling in Cloud Computing Environment)

11EMERGING TECHNOLOGIES AND INNOVATION
(Source: Novak, Kasera and Stutsman 2019)
3.1 Virtual Cluster Monitor System:
The virtual cluster monitor system is a specific type of system which is capable of
detecting whether active sessions of HTTP are over the threshold or not within a virtual cluster.
For the computing tasks that are distributed in nature, the virtual cluster monitor system is
capable of detecting whether the virtual machine numbers are over the threshold of using the
physical type of resources within a virtual cluster [13]. The virtual cluster monitor system is
mainly used for controlling and triggering the scaling down and scaling down within auto
provisioning system. This is done on the virtual machine instances number depending on the
scaling indicator statistics. The virtual cluster monitor system also collects data regarding
resource utilization of all the virtual machines for each of the virtual cluster operating in a cloud
computing environment. In the aspects of virtual cluster monitor system, the virtual clusters
consist various of same server that are divided up the same way. For handling the more intense
and critical jobs more virtual instances might need to be included within the workflow. A virtual
cluster can be formed using several of virtual machines through networking.