a survey on applications of cloudlet infrastructure in mobile

International Journal of Latest Trends in Engineering and Technology
Vol.(8)Issue(1), pp.309-318
DOI: http://dx.doi.org/10.21172/1.81.040
Srilatha.M1, S.Rajeshwari2 and K.Rani3
Abstract:Mobile cloud computing (MCC) integrates the cloud computing into the mobile environment to
improve the performance (e.g., battery life, storage, and bandwidth), environment (e.g., scalability, and
availability), and security (e.g., reliability and privacy). The cloudlet is one of the mobile cloud computing
infrastructures to reduce power and network delays. In this paper we presented different application and research
proposals of cloudlets.
Key words: Mobile cloud computing, cloudlet.
MCC is emerging research area that combines Cloud Computing technology and the mobile
environment to create new infrastructure that is capable of performing intensive computations
and storing massive amount of data[1]. The mobile device users are provided with an online
access to an unlimited computing power and storage space as shown in fig 1.
Fig 1.
Cloudlet is resource-collection of computers which is well-connected to the Internet and
available for use by nearby mobile devices. Hence, when mobile devices cannot or do not want
Deaprtment of Computer Science, VR Siddhartha Engineering College Andhrapradesh , India
Deaprtment of Computer Science, VR Siddhartha Engineering College Andhrapradesh , India
Deaprtment of Computer Science, VR Siddhartha Engineering College Andhrapradesh , India
A Survey On Applications Of Cloudlet Infrastructure In Mobile Cloud Computing
to connect to the cloud, they can find and access a nearby computing resource. In this way
mobile device offloads its workload to a local ‘cloudlet’ comprised of several multi-core
computers with connectivity to the remote cloud servers. . A cloudlet is similar to a small data
center that is located on designated areas/places and is connected to a larger cloud server via the
Cloudlets refer to the infrastructure consisted of local physical equipments attach to different
department or unit, inside the cloud computing based information system [5]. These
infrastructures are distributed on location but stay at the same level logically as shown in figure
Fig 2.
In paper [2] e presented efficient techniques to reduce energy consumption in MCC using
Offloading, green MCC, Cloudlet architecture, and clone.
In paper [3] authors are implemented MOCHA (mobile-cloudlet-cloud architecture) for face
recognition by partitioning the computation (tasks) among the available cloud servers and the
cloudlet. This can be done in teo ways (1) Fixed: the tasks are equally distributed among the
available cloud servers (or the cloudlet). (2) Greedy: based on cloud servers response times task
are allocated to the server (cloudlet) that can complete the task in the minimum amount of time.
Authors implemented Cloud-Vision face recognition application perform the face recognition for
a mobile device.It is is executed in two separate phases: 1) face detection using Viola-Jones
algorithm that progressively narrows an initial large set of face location candidates to the final
set of detected face locations; and 2) face recognition, which uses the classic Eigenfaces
approach, to determine the likelihood of each detected face matching one of the template images
in a database. These computation can be partitioned among mulitple cloud servers to speed up
the response time.
In paper [4] focused on an optimal admission control policy for MCC. There is no standard
mechanism to charge users based on their usage of services by service providers. Hence authors
proposed an optimization model for allocating resources of bandwidth and cloudlet to meet the
requirement of mobile users based on the semi-Markov decision process (SMDP). Here the
Srilatha.M, S.Rajeshwari and K.Rani
policy of SMDP obtained by transforming an original optimization problem into a linear
programming (LP). In this processes an admission control mechanism of the MCC hotspot is
designed to decide whether an arriving request from mobile user can be accepted or not. It takes
the state of MCC hotspot into account and makes decision according to the optimal policy to be
obtained from SMDP.
When Given a set of users operating battery-powered devices and requesting different types of
cloud services and a set of cloudlets then power consumption may be more. In paper [6]
suggested a solution to assign each user request to a specific cloudlet capable to service by
minimizing the overall power consumption. The power consumption was minimized by
examining that whether the user is in coverage region of cloudlet at particular time slot or not
and then applied a cost function to represent total power consumption of the system.
In paper [7] focused on impact of cloudlets on interactive mobile cloud applications (IMCA)
such as file editing, video streaming and chatting with mobile nodes moving from one cloudlet
coverage to another cloudlet coverage area. To evaluate performance in IMCA authors are
implemented a hybrid ad-hoc-like wireless routing protocol by modifying the DestinationSequenced Distance-Vector routing (DSDV) ad-hoc routing protocol in NS2. Then compared
experimental result with pure cloud computing communication. Finally concluded that using
cloudlet data transfer delay will be reduced and throughput will be increased.
In article [8] specified that a mobile user exploits virtual machine (VM) technology to customize
service software on a nearby cloudlet (like in fig.) and then uses that service over a wireless
LAN. To deploy cloudlet like wi-fi it is needed to simplify cloudlet management for integrating
Wi-fi with cloudlet .In order to do this, authors introduced a solution called transient
customization of cloudlet infrastructure using hardware VM technology. It emphasis on pre-use
customization and post-use cleanup ensure that cloudlet infrastructure is restored to its pristine
software state after each use, without manual intervention.
[9] In paper presented a novel cloudlet-based efficient data collection system in Wireless Body
Area Networks (WBANs). Wireless Body Area Networks (WBAN) consists of a group of
communicating sensor nodes, which can monitor different body parameters and gather a lot of
body information. These devices, communicating through wireless technologies, can transmit
data from the body to a base station from where the data can be forwarded to a hospital, clinic, or
a service provider in real-time manner as shown in fig 3. The huge amount of data collected by
WBAN nodes demands scalable, on-demand, powerful, and secures storage and processing
Fig 3.
A Survey On Applications Of Cloudlet Infrastructure In Mobile Cloud Computing
So authors proposed a prototype of WBANs, including Virtual Machine and Virtualized
Cloudlet. It had been evaluated using extended CloudSim simulator. Using this prototype users
or service providers can accesses data by minimizing end-to-end packet cost and packet–to-cloud
In paper [10] proposed an analytical approach for performance analysis and modeling of cloudlet
called interacting stochastic sub-models. In this processes, different servicing steps of cloudlet
architecture mapped into separate stochastic sub-models. These sub-models interact with each
other such that the outputs of ones are transmitted to the other as inputs and vice versa. By
iteration over this process the final results will be obtained. These stochastic sub-models are
named as Request Admission Engine(RAE), Host selection engine(HSE), Cloudlet provisioning
engine (LPE) and public cloud provisioning Engine (PPE)) that each of which covers the
behavior of appropriate queue and its corresponding engine.
In paper [11], proposed pocket cloudlets, an effective architecture that leverages abundant nonvolatile memories(NVM )in mobile devices to significantly improve user experience, both in
terms of latency and battery life, by avoiding expensive radio wakeups and transmissions to
access cloud services. The problems with mobile cloud services are like the amount of data to be
stored locally on the device, a mechanism to manage the locally stored cloud data, storage
architecture for efficiently storing and accessing this large amount of data.
To solve these problems authors implemented PocketSearch, a mobile search pocket cloudlet
that lives on the phone and is able to answer queries locally without having to use the 3G link2.
It consists of two components: the community and the personalization components. The
community part of the cache is responsible for storing the small set of queries and search results
that are popular across all mobile users. This information is automatically extracted from the
search logs and is updated overnight every time the mobile device is recharging. The
personalization part of the cache monitors the queries entered as well as the search results
clicked by the user and performs two discrete tasks. First, it expands the cache to include all
those queries and search results accessed by the user that did not initially exist in the community
part of the cache. Second, it collects information about user clicks, such as when and how many
times the user clicks on a search result after a query is submitted, to customize ranking of search
results to user’s click history. When a query is submitted, PocketSearch will first perform a
lookup in the cache to find out if there are locally available search results for the given query. In
the case of a cache hit, the search results are fetched from the local storage, ranked based on the
past user access patterns recorded by the personalization part of the cache, and immediately
displayed to the user. In the case of a cache miss, the query is submitted to the search engine
over the 3G radio link.
In paper [12], authors analyzed the computation offloading problem in cloudlet-based mobile
cloud computing. They worked on the computation offloading strategy of multiple users via
multiple wireless APs using a game-theoretic solution. It uses two algorithms with respect to
heterogeneous and homogenous mobile users. The performance of this algorithm is compared
Srilatha.M, S.Rajeshwari and K.Rani
with two other computation offloading strategies:-1) Local Computation: all mobile users
execute their computation tasks locally on their mobile devices. 2) Random Selection: mobile
users either randomly chooses one of the wireless APs to offload their computation tasks or
execute their tasks locally.
In paper[13] , authors proposed an advanced dynamic model, dynamic energy-aware cloudletbased mobile cloud computing model (DECM), that leverages cloudlets technique to assign,
manage, and optimize the cloud-based infrastructure usages and services for achieving green
computing. This model uses dynamic programming to assist cloudlets to determine the cloud
computing resources within a changing operational environment. Where cloud lets are
conceptualize as a dynamic cloudlet (DCL) to achieve following missions:
Determining and predict physical machines on the cloud using dynamic programming
Calculations of simple applications and quickly respond to end users.
Predict whether users should switch to other cloudlets and enable a real-time suggestion
for cloudlets switching.
The scenario of this algorithm is that the mobile device connects to the cloud server by one route
M1 at the beginning of the service implementation. Moving forward to the next time unit, a
decision will be made depending on the comparison among all routes. The route(s) with a better
performance will be selected. The resolution will be made for each time unit by using the same
method and finally reach a minimum total energy cost with quality services in a defined timing
period. This approach was efficient in a specific condition that implies the users can accept a
scalable service delivery manner during the usages.
In paper [14] proposed a cloudlet-based multi-lateral resource exchange framework for mobile
users, relying on no central entities. In this paper proposed a generic mobile cloud computing
framework (with underlying technology BitCoin) where all types of underlying resources can be
exposed to the mobile users through virtualization. According to this processes a Cloudlet is
deployed by a user provide services to the other mobile users as shown in fig 4.
Fig 4.
A Survey On Applications Of Cloudlet Infrastructure In Mobile Cloud Computing
To check the effectiveness of this framework, authors were also implemented a prototype
enabling Internet bandwidth leasing among mobile user through the users cloudlet servers. This
had done by splitting energy-consuming tasks to the cloudlet side, while keeping necessary userinteractive tasks on the mobile side.
The connections between a mobile user and mobile cloudlets can be intermittent because of
mobility and cloudlet capacity. As a result, offloading actions taken by the mobile user may fail.
So in paper [15], authors developed an optimal offloading model Markov decision process
(MDP) based dynamic offloading algorithm for the mobile user to intermittently connect cloudlet
system, by considering the users’ local load and availability of cloudlets. The algorithm is
designed as follows
During the application execution, assumed a time interval t for the user to finish
executing any application phase locally or remotely on cloudlets.
At the beginning of each decision period, the user observes the current system states, i.e.,
the number of jobs Q in the queue, the application phase G that the user is executing, and the
number of accessible cloudlets N .
Based on the observed composite state S = (G, Q, N ) of the system, the user computes
the immediate costs of local execution and offloading
In General cloudlets have limited resources and processing abilities, which implies that they may
not be capable to process every incoming request. Instead, some resource-intensive requests need
to be sent to remote data centers for processing. In paper [16] addressed the online request
admission issue in a cloudlet by maximizing the system throughput. The online request
throughput maximization problem is to determine whether an arrival request to be admitted or
rejected by the system such that the system throughput is maximized for a specified time period
T. The system throughput is the ratio of the number of admitted requests to the number of
requests for period T. The algorithm for online request throughput maximization is processed as
It first checks whether the requested amount of each resource r i , k (t) can be met by
the system.
If not, the request is rejected immediately;
Otherwise the system calculates the admission cost of processing the request based on
the load of each resource at this moment.
If its admission cost is beyond a specified threshold of each resource in the system,
the request will be rejected; otherwise, it is admitted by the cloudlet.
In paper [17] proposed a mechanism to identify a cloudlet for computation offloading in a
decentralized manner. It will be dine in two ways.
The first phase Mobile device acquires information of the available cloudlets
The mobile device now has to select a cloudlet based on the information and requirement
of the application to be offloaded.
Srilatha.M, S.Rajeshwari and K.Rani
If the application requirement related to RAM and CPU is known beforehand, cloudlet
specifications given in the Cloudlet structure related to memory and CPU are compared. If the
cloudlet doesn’t have enough memory, the application may not run with proper performance and
hence should not be chosen.
In paper [18] compared the performances of translating various n-grams executed on a cloud and
on a cloudlet by using multilingual dictionary application based on the dynamic VM synthesis. It
consists of the following elements:
A server on which the mobile device sends words or phrases for translation
Translation of the words or phrases, and returning appropriate translation back to the
mobile device.
If cloudlet is set as a cluster structure, VMs can be easily cloned and thus gain parallelism. The
processes of dynamic VM synthesis are shown in fig 5.
Fig 5
When customer tries to translate any word or phrase, first the mobile device will try to find if
there are local LAN cloudlets available. The mobile device is connected to the cloudlet by using
high-bandwidth wireless LAN. If there is no cloudlet which can perform such tasks then the
mobile application tries to connect to a distant WAN based cloud server which can perform the
In paper[19] presented Personal Cloudlets framework called OPENi’s to allows users share,
reuse, and control access to their data across many mobile applications by maintaining cloud
scalability. OPENi consist of two APIs:
The API framework: allows for frictionless interoperability between cloud based
The Personal Cloudlet framework is a virtual space that securely stores user data and
gives users primary control over that data.
Along with above two components it also consists of the other components called mobile client
library, the security framework. The architecture of OPENi was described as follows
A Survey On Applications Of Cloudlet Infrastructure In Mobile Cloud Computing
The main objective of this approach is to:
1) To build a web based security and authorization framework that will satisfy the service
provider’s requirements and a context broker that will enable the sharing of context information
between applications in accordance with the users privacy settings.
2) To deliver an open source platform that will allow application consumers to create, deploy
and manage their personal space in the cloud (Personal Cloudlet). Each Personal Cloudlet will
constitute a novel entity that will be linked to its user’s identity over the web in a similar way
that a social profile does today.
3) To provide and promote a novel, user-centric application experience of cloud-based services
not only across different devices but also inherently across different applications. The OPENi
framework will enable application consumers to share and distribute their data across their
4) To ensure the OPENi platform maintains a low barrier to entry for application developers and
service providers.
In paper [20] described a distributed system that creates a 3D map of the complete world that is
continuously updated by crowd-sourced depth information. In this model, information of system
is encoded as a depth map (or point cloud): a set of discrete points with 3D coordinates. Then the
point cloud spited into sub models when it has become too large to ensure fast and accurate
registration. Now each submodel is assigned to a different VM. The submodels hosted in the
VMs on a cloudlet will contain details of geographically close environments. Users are
connected to a nearby cloudlet, the submodel is deployed on the cloudlet to which the user was
connected. A central locator component keeps the mapping between geographical coordinates
and cloudlet IP.
Srilatha.M, S.Rajeshwari and K.Rani
[1] Niroshinie Fernando, Seng W. Loke , Wenny Rahayu,“Mobile cloud computing”: A survey.In Fututre
Genetation Computer System 29(2013) 84-206.
[2] Khadijah S. Bahwaireth , Lo’ai Tawalbeh , Anas Basalamah, Yaser Jararweh Mohammad, Tawalbeh, “Efficient
Techniques for Energy Optimization in Mobile Cloud Computing”: 978-1-5090-0478-2/15/$31.00 ©2015 IEEE.
[3] Soyata T, Muraleedharan R, Funai C, Kwon M, Heinzelman W. Cloud-vision, ”Realtime face recognition using
a mobile-cloudlet-cloud acceleration architecture”: IEEE symposium on computers and communications (ISCC).
IEEE; 2012.p.059–66.DOI: http://dx.doi.org/10.1109/ISCC.2012.6249269.
[4] Hoang DT, Niyato D, Wang P,”Optimal admission control policy for mobile cloud computing hotspot with
cloudlet”: IEEE wireless communications and networking conference (WCNC)” IEEE;2012. p. 3145– 9. DOI:
[5] Han Xingye, Li Xinming, Liu Yinpeng,“ Research on Resource Management for Cloud Computing Based
Information System”: 2010 International Conference on Computational and Information Sciences.
[6] M. Al-Ayyoub, Y. Jararweh, L. Tawalbeh, E. Benkhelifa, and A.Basalamah,“Power Optimization of Large
Scale Mobile Cloud Computing Systems”:IEEE Fi-Cloud, 3rd International Conference on Future Internet of Things
and Cloud, Rome, Italy, 24-28 Aug, (2015).
[7] Fesehaye D, Gao Y, Nahrstedt K, Wang G, “Impact of cloudlets on interactive mobile cloud applications”:
IEEE 16th international enterprise distributed object computing conference (EDOC). IEEE; 2012. p. 123–32. DOI:
http://dx.doi.org/10. 1109/EDOC.2012.23.
[8] Satyanarayanan M, Bahl P, Caceres R, Davies N,”The case for vm-based cloudlets in mobile computing”: IEEE
Pervasive Compute 2009;8(4):14- 23.http://dx.doi.org/10.1109/MPRV.2009.82.
[9] Quwaider M, Jararweh Y.,” Cloudlet-based efficient data collection in wireless body area networks”:
Simulation Model Practice and Theory 2015; 50:57–71. http://dx.doi.org/ 10.1016/j.simpat.2014.06.015.
[10] Raeia , Nasser Yazdani ,Reza Shojaee, “Modeling and Performance Analysis of Cloudlet in Mobile Cloud
[11]Emmanouil Koukoumidis, Dimitrios Lymberopoulos,”Pocket Cloudlets”: Karin Strauss.ASPLOS’11, March 5–
11, 2011, Newport Beach, California, USA, ACM 978-1-4503-0266-1/11/03.
[12] Xiao Ma, Chuang Lin, Xudong Xiang” Game-theoretic Analysis of Computation Offloading for Cloudletbased Mobile Cloud Computing”: MSWiM’15, November 2–6, Cancun, Mexico.
[13] Gai K, et al, ”Dynamic energy-aware cloudlet-based mobile cloud computing model for green
computing”: Journal of Network and Computer Applications (2015), http://dx.doi.org/10.1016/j.jnca.2015.05.016i.
[14] Y. Wu, L. Ying,”A Cloudlet-based Multi-lateral Resource Exchange Framework for Mobile Users”:
proceedings of the IEEE INFOCOM, 2015, pp. 927–935. DOI:10.1109/INFOCOM.2015.7218464.
[15] Y. Zhang, D. Niyato, W. Ping, “Offloading in mobile cloudlet systems with intermittent connectivity”:IEEE
Transactions on Mobile Computing 14 (12) (2015) 2516–2529.
A Survey On Applications Of Cloudlet Infrastructure In Mobile Cloud Computing
[16] ] Q. Xia, W. Liang and W. Xu, “Throughput maximization for online request admissions in mobile cloudlet”
:IEEE 38th Conference on Local Computer Networks (LCN 2013), October 2013.
[17] Dilay Parmar, A. Sathish Kumar, Ashwin Nivangune, “Discovery and Selection Mechanism of Cloudlets in a
Decentralized MCC environment”: 2016 IEEE/ACM International Conference on Mobile Software Engineering and
[18] Aleksandar Bahtovski, Katerina Zdravkova, Marjan Gusev,“Performance of Cloudlet-based
[19] D´onal McCarthy , Paul Malone, etc, “Personal Cloudlets: Implementing a User-Centric Data store with
Privacy Aware Access Control for Cloud-based Data Platforms“:2015 IEEE DOI: 10.1109/TELE RISE.2015.15.
Download PDF