Femtolet: A novel fifth generation network device for green mobile cloud computing

Simulation Modelling Practice and Theory 62 (2016) 68–87

Contents lists available at ScienceDirect

Simulation Modelling Practice and Theory journal homepage: www.elsevier.com/locate/simpat

Femtolet: A novel fifth generation network device for green mobile cloud computing Anwesha Mukherjee∗, Debashis De Department of Computer Science and Engineering, West Bengal University of Technology BF-142, Salt Lake, Sector–I, Kolkata-7000064, India

a r t i c l e

i n f o

Article history: Received 27 October 2015 Revised 31 December 2015 Accepted 22 January 2016

Keywords: Cloudlet Femtocell Femtolet Latency Power Green

a b s t r a c t For the first time, this paper proposes a new network device denoted as ‘Femtolet’ for fifth generation mobile network to provide communication and computation offloading facilities simultaneously at low power and low latency. The features of two separate devices femtocell base station and cloudlet are combined into a single network device denoted as Femtolet to provide the services of femtocell along with a cloud environment at low power and low latency. The architecture and working principle of the proposed device Femtolet are discussed with its power consumption model and latency. Mathematical analyses present that using Femtolet instead of the existing femtocell plus cloudlet and Small Cell cloud-enhanced e-node B architectures, power consumption can be reduced by approximately 17% and 11% respectively. Mathematical analyses also show that the proposed device Femtolet can reduce the latency by approximately 20% and 13% than the existing femtocell plus cloudlet and Small Cell cloud-enhanced e-node B architectures respectively. The proposed working model of Femtolet is simulated using Qualnet7 and its performance is analyzed with respect to average delay, average jitter, energy consumption in transmit and received modes, carried load, throughput, offloading time and offloading power. © 2016 Elsevier B.V. All rights reserved.

1. Introduction According to Cooper’s law the number of voice or data conversations conducted in all radio spectrum on a given region doubles every 2.5 years.1 Cooper’s law is pictorially demonstrated in Fig. 1. With this exponential growth in the number of mobile users, network densification, power efficiency and low latency have become critical issues in the area of fifth generation mobile network [1–3]. Fifth generation (5G) mobile network has to offer high signal strength and high data rate at the same time. To provide high signal strength especially at indoor region, the femtocell technology has been developed as home node base station (HNB) [4–7]. On the other hand the demands of mobile web users to run heavier applications are increasing day by day. But the mobile phone faces some difficulties like small storage space, limited processing power, limited battery life etc. To satisfy the user demands by overcoming these problems, mobile cloud computing (MCC) has been introduced as a combination of mobile computing and cloud computing [8–11]. Cloud is the combination of virtualization of large amount of resources with a distributed computing paradigm integrated with Software as a Service (SaaS), Platform as a Service (PaaS) and Infrastructure as a Service (IaaS) [8–12]. In MCC the data processing and data storage both take ∗

1

Corresponding author. E-mail addresses: [email protected] (A. Mukherjee), [email protected] (D. De). http://www.comsoc.org/ctn/will- densification- be- death- 5g.

http://dx.doi.org/10.1016/j.simpat.2016.01.014 S1569-190X(15)30159-3/© 2016 Elsevier B.V. All rights reserved.

A. Mukherjee, D. De / Simulation Modelling Practice and Theory 62 (2016) 68–87

69

Fig. 1. Graphical representation of Cooper’s law of spectral efficiency.

place outside the mobile device (MD) and into the cloud. MCC provides all the cloud services and applications like mobile commerce, mobile healthcare, mobile gaming etc. But long distance cloud access affects the quality of experience (QoE) of the users by increasing the wireless area network (WAN) delay. To provide high bandwidth wireless access to the cloud, cloudlet is introduced in the network [13,14]. A cloudlet is a reliable, resource-rich computer or number of computers connected to the Internet and accessible to the nearby MDs. It contains cache copies of the data already available inside the cloud. Accessing cloudlet through femtocell provides low latency but high bandwidth and secure wireless access [4,9]. To fulfill the aim of 5G mobile network, Small Cell cloud-enhanced e-node B (SCceNB) is introduced which provides communication with additional computation facility [15–17]. These small cell nodes are connected with a small cloud which is referred as femtocloud. The femtocloud has an intermediate storage and computation ability. The femtocloud is connected to the core cloud. Our aim is to reduce the latency as well as power consumption even than SCceNB. Hence in this paper we have proposed a new network device denoted as ‘Femtolet’ for 5G mobile network. 1.1. Motivations and contributions of proposed work In femtocloud based 5G mobile network cloud service access at low latency is achieved either using (femtocell + cloudlet) or using SCceNB scenario. Accessing the cloudlet instead of the long distance cloud provides low latency i.e. high speed Internet access. But if the cloudlet is unable to provide the required service, then the cloud is accessed. In such a case due to the communication between the femtocell and cloudlet, additional power consumption and delay are introduced. This affects the QoE. On the other hand, the SCceNB provides limited cloud functionalities. If the SCceNB is unable to satisfy the user’s demand, the femtocloud serves the user. If the femtocloud is also unable to satisfy the user’s need, then the cloud is accessed. As the communication takes place between the SCceNB and femtocloud, and between the femtocloud and cloud, additional power consumptions and delays are introduced. As a result the QoE gets affected. To deal with this problem, our motivation is to propose a new network device for 5G mobile network which will provide communication as well as computation at low power consumption and low latency. The contributions of this paper are: (i) A new network device referred as ‘Femtolet’ is proposed for 5G mobile network by incorporating the features of both femtocell and cloudlet. The architecture and working model of Femtolet are discussed. Femtolet can serve as a home base station like femtocell as well as can act as a cloudlet for offloading data and applications to save battery life of the MDs registered under it. (ii) The proposed network device denoted as Femtolet provides: • Power saving with respect to the existing (femtocell + cloudlet) and SCceNB scenarios. • Latency reduction with respect to the existing (femtocell + cloudlet) and SCceNB scenarios. 1.2. Organization of paper This paper is organized as follows: Section 2 presents the related works, Section 3 describes the architecture and working principle of proposed network device denoted as Femtolet; the power and latency consumption models of Femtolet are developed in Section 4; performance analysis of Femtolet is discussed in Section 5; the research challenges for Femtolet are discussed in Section 6; finally conclusion is given in Section 7.

70

A. Mukherjee, D. De / Simulation Modelling Practice and Theory 62 (2016) 68–87

2. Related works Femtocell is an interesting evolution in the world of green mobile communication technology. Femtocell is generally allocated inside a macrocell or a microcell in order to offer good indoor coverage. Femtocell is known as home node base station because it is usually deployed inside office buildings or residential houses to provide better connectivity to the network. The coverage region of a femtocell base station is 10–20 m. Because of small coverage area, the power transmission of a femtocell base station is low. In our previous works it has been demonstrated that using femtocell green mobile network can be developed [4,5]. The architecture of femtocell has been discussed in [18]. Cloudlet is a recent development in the field of MCC. To reduce the delay of remote cloud access, cloudlet is developed. A cloudlet is referred as a group of computers having gigabit connectivity and high-bandwidth wireless access to the network [13,14]. A cloudlet contains cache copies of code and data available inside the cloud. A cloudlet provides cloud services to its nearby MDs. As cloudlet is at a close distance from the MD, the latency is reduced. The QoE is improved accordingly. Cloudlet is usually deployed inside a building. The MDs located inside the building can offload data as well as computation to the cloudlet instead of the remote cloud. As a result the delay is reduced. In [19] an overview of MCC has been given. In [20–22] application offloading using MCC has been discussed. In [23] a simulation framework has been discussed for MCC [24]. Cloudlet [25] based hierarchical storage for MCC environment has been illustrated in [26]. Femtocell based cloud access has been discussed in [27–29] and the network has been referred as femto-cloud network. Interference management for macro-femtocell and micro-femtocell based mobile network has been discussed in [30]. Fifth generation mobile network is going to be launched in 2020. 5G mobile network integrates radio and computational resources to provide high signal strength, high bandwidth, and high speed Internet connectivity. SCceNB has been introduced [15,16] as small cell base stations with additional cloud functionalities to offer high speed cloud access. 5G mobile network deals with dense deployment of small cells with short range communication. To offer high capacity and high speed connectivity, millimeter wave link is used in 5G [16]. This link provides high speed offloading than traditional backhaul link. In 5G mobile network millimeter wave of 28–38 GHz is widely used though higher frequencies also can be used in 5G. In order to enhance the spectral efficiency, massive multiple-input, multiple-output transceivers are used in 5G mobile network. This in turn reduces the time of offloading computation from MD to the cloud. Although the existing small cell base station with short range communication provides good connectivity, but remote cloud access introduces delay. To reduce latency, cloudlet can be used. For small cell base station with additional cloud functionalities i.e. SCceNB, femto-cloud provides a medium storage and computation facility. In both of these cases, communication delay occurs either between small cell base station and cloudlet or between SCceNB and femto-cloud. To avoid this problem, in our paper Femtolet is proposed as a small cell base station possessing storage and computational capabilities.

3. Architecture and working principle of proposed network device denoted as Femtolet 3.1. Proposed architecture of Femtolet The proposed architecture of Femtolet is demonstrated in Fig. 2. The components of femtocell [18] and cloudlet [13,14] are integrated to produce Femtolet. Femtolet can be deployed inside a macrocell or microcell to provide good indoor coverage

Fig. 2. Proposed architecture of Femtolet integrating the components of femtocell and cloudlet.

A. Mukherjee, D. De / Simulation Modelling Practice and Theory 62 (2016) 68–87

71

like femtocell as well as to offer cloud services like cloudlet. The main architectural difference of Femtolet from femtocell is that: femtocell does not provide the cloud services but Femtolet provides the cloud services such as IaaS, PaaS and SaaS. Functionalities of the components of Femtolet: (i) Field Programmable Gate Array (FPGA) manages hardware authentication, data encryption and network time protocol. (ii) Microprocessors manage radio protocol stack and associated baseband processing. (iii) Random Access Memory (RAM) handles information associated with user mobility, network traffic and interference condition. RAM is associated with FPGA and microcessor. (iv) Radio Frequency transmitter (RFT), Radio Frequency receiver (RFR) and Power Amplifier (PA) perform other nonessential functions. (v) A cloud platform is provided with an integration of storage, applications and services like IaaS, PaaS and SaaS: (a) In IaaS, storage and computational capabilities are supplied as a consistent service over the network in an ondemand fashion. IaaS providers offer a variety of resources to perform high performance computing applications. (b) PaaS incorporates operating system (OS), database, web server and programming language execution environment to afford a cloud platform for deploying various applications. (c) In SaaS, different applications software are provided and the software run on the cloud. The cloud user accesses the software as a service. 3.2. Proposed working model of Femtolet The working model of our proposed network device denoted as Femtolet is shown in Fig. 3. Maximum user capacity and deployment of Femtolet are similar to that of femtocell [4,18]. Resource allocation and management in Femtolet is same as cloudlet [12–14]. From Fig. 3 it is observed that different types of MDs like laptop, tablet, mobile phone, android phone are connected with Femtolet through the Uu interface. Uu interface is an external interface that connects the base station (here Femtolet) with the user equipment (here mobile device). This interface is used in Universal Terrestrial Radio Access Network (UTRAN). This interface connects the user equipment with UTRAN network and known as Uu interface. Femtolet operates in licensed spectrum and connects the MDs under its coverage to the operator’s network through the use of residential DSL or cable broadband connection. It contains an internal storage to store large amount of data currently accessed by the MDs registered under the Femtolet. Femtolet is connected to the HNB-gateway (HNB-GW) through the Internet security gateway via the Iuh interface. The HNB-GW is connected to the core network via the Iucs/Iups interface. The Iu interface is used to connect the Radio Network Controller with the Core Network. Femtolet can have a coverage area of 10–20 m like femtocell. The registration and deregistration of Femtolet and MD are performed by the home nodeB application part (HNBAP) protocol [18]. The working principle of Femtolet is presented as follows: • Femtolet registration: Femtolet sends a registration request message containing the Femtolet ID and Femtolet location to the HNB-GW. After receiving the message, if the HNB-GW sends registration accept message to the requested Femtolet, the registration process is completed. Else if the HNB-GW is unable to accommodate a new Femtolet, it sends registration reject message to the Femtolet, and registration is not performed. • Femtolet deregistration: Femtolet can deregister from the HNB-GW by sending Femtolet deregistration message containing the reason of leaving. By this message the Femtolet is automatically deregistered from the HNB-GW. • MD registration: A Femtolet can accommodate a fixed number of MDs. When a MD comes under its coverage, the Femtolet initiates registration of that MD by sending MD registration request message containing the MD ID to the HNB-GW. If the HNB-GW sends registration accept message containing the Iu connection context ID to the requested Femtolet, the MD is successfully registered under it. Else if the Femtolet is already fully loaded i.e. serving maximum number of MDs which it can accommodate, the HNB-GW sends registration reject message, and the MD is not registered. • MD deregistration: The MD can deregister from the HNB-GW by sending MD deregistration message containing the context ID. At this time the MD connection context ID is deleted and the MD is automatically deregistered from the HNB-GW. • Serving as femtocell: Femtolet is connected to the core network. The MDs registered under a Femtolet can generate or receive call and send messages. In this case the Femtolet acts as a femtocell. • Serving as cloudlet: Femtolet is connected to the Internet and contains a large storage. It stores the most frequently accessed data from the cloud. The Femtolet first accesses the cloud and then temporarily stores the cache copies of data and code available in the cloud. MDs registered under the Femtolet can access the data stored by that Femtolet. A cloud platform is provided with an integration of storage, applications and services. If a user asks for information not available inside the Femtolet, the Femtolet asks the cloud. After receiving the data from cloud, the Femtolet stores the copy within its storage and serves the user. A MD registered under the Femtolet can offload data and computation both inside the Femtolet to access cloud services at low latency. • Secure information management: A Femtolet is connected to the network via a security gateway (Se-GW) for secure data transmission. For security purpose the following precautions are to be taken in Femtolet [9,13]: • Use of adequate cryptographic algorithms for authentication, confidentiality and data integrity. • Prevention of disclosing MD IDs registered under a particular Femtolet. • Make the sensitive data such as user information, authentication details, user and control plane data inaccessible in plain text at the Femtolet.

72

A. Mukherjee, D. De / Simulation Modelling Practice and Theory 62 (2016) 68–87

Fig. 3. Proposed working model of Femtolet with registration and deregistration procedures.

• • • •

Authentication of the Se-GW by the Femtolet. Authentication of the Femtolet by the Se-GW. Permitting the Femtolet to access the core network only if all required authentications are fulfilled successfully. For safe deployment, a tamper-resistant or tamper-evident enclosure should be contained in the Femtolet with thirdparty remote monitoring of hardware integrity. • For securing the data access inside the Femtolet, biometric authentication e.g. keystroke analysis, finger print analysis etc. can be used. • In case of key-stroke analysis, the keystroke of a user is verified to check whether he/she is authentic user or not. If the user is authentic, he/she can access the cloud services of Femtolet. Else he/she is not allowed. If finger print analysis is used, the finger print of the user is verified to check whether he/she is an authentic user or not. If the user is authentic, he/she is able to access the cloud services offered by Femtolet. Else he/she is not allowed. In this way, using bio-metric authentication unauthorized access to the cloud services of Femtolet can be prevented. 3.3. Life cycle model of Femtolet Fig. 4 shows the life cycle model of Femtolet. As shown in Fig. 4, there are four states: • • • •

State State State State

1: 2: 3: 4:

Start (Femtolet is turned on) Active (Femtolet becomes active and serves users under its coverage) Sleep (Femtolet goes to idle mode when no user is present under its coverage) Inactive (Femtolet is turned off)

A. Mukherjee, D. De / Simulation Modelling Practice and Theory 62 (2016) 68–87

73

Fig. 4. Life cycle model of Femtolet.

The Femtolet is first turned on which is referred as ‘Start’ i.e. state 1. Then the Femtolet moves to state 2 i.e. ‘Active’. From state 2 Femtolet can move to state 3 i.e. ‘Sleep’. From state 3 the Femtolet can be turned off and then it goes to state 4 i.e. ‘Inactive’. When a Femtolet is turned on, it becomes active. In active state Femtolet can provide service to the users under its coverage. When no user is present inside the coverage of a Femtolet, it goes to idle mode which is referred as sleep state. In this state, Femtolet can receive information. When a user enters into its coverage, it goes back to active state. When the Femtolet is turned off, it goes to inactive state. As the probability of occurrence of each state is dependent on the immediate previous state, Markov Chain model is used to determine the state transition probability [31,32].2 If s is a state sequence and l is the length of the sequence, it is given as (s1 s2 , .... , sl ), where each state is denoted by si and si  s.

2

http://pages.cs.wisc.edu/∼molla/summer_research_program/lecture5.1.pdf.

74

A. Mukherjee, D. De / Simulation Modelling Practice and Theory 62 (2016) 68–87

According to the Markov chain model, the probability of occurrence of a state depends on its previous state, given as,

P ( s ) = P ( s1 )

l 

P (si |si−1 )

(1)

i=2

Let S1 , S2 , S3 and S4 present state 1, state 2, state 3 and state 4 of Femtolet respectively. Then the probability of occurrence of each state is determined using the Laplace estimation and Bayesian approach, as follows,

num1 + x p1 P ( S1 ) =  num j + x

(2)

j

num2 + x p2 P ( S2 ) =  num j + x

(3)

j

num3 + x p3 P ( S3 ) =  num j + x

(4)

j

num4 + x p4 P ( S4 ) =  num j + x

(5)

j

where num1 , num2 , num3 and num4 present the number of occurrence of state S1 , S2 , S3 and S4 respectively, p1 , p2 , p3 and p4 present the prior probability of the occurrence of state S1 , S2 , S3 and S4 respectively, numj denotes the number of occurrence of state j and j is not a terminating state, and x denotes the number of virtual instance. As there are four states, the uniform probability of occurrence of each state is 0.25. To determine the probability of state transition, state sequence is needed. Let (S1 S2 S3 S2 S3 S4 ) be a state sequence. Then the probability of state transition considering the sequence is given as,

P ( S1 |S1 ) =

0.25x 1 + 0.25x 0.25x 0.25x , P ( S2 |S1 ) = , P ( S3 |S1 ) = , P ( S4 |S1 ) = , 1+x 1+x 1+x 1+x

P ( S1 |S2 ) =

0.25x 0.25x 2 + 0.25x 0.25x , P ( S2 |S2 ) = , P ( S3 |S2 ) = , P ( S4 |S2 ) = , 2+x 2+x 2+x 2+x

P ( S1 |S3 ) =

0.25x 1 + 0.25x 0.25x 1 + 0.25x , P ( S2 |S3 ) = , P ( S3 |S3 ) = , P ( S4 |S3 ) = , 2+x 2+x 2+x 2+x

P (S1 |S4 ) = 0.25, P (S2 |S4 ) = 0.25, P (S3 |S4 ) = 0.25, P (S4 |S4 ) = 0.25. As S4 is a terminating state, the probability of state transition from S4 is set to uniform probability of occurrence i.e. 0.25. The state transition diagram with probabilities is presented in Fig. 4 where S1 , S2 , S3 and S4 present ‘Start’, ‘Active’, ‘Sleep’ and ‘Inactive’ state respectively. As in our problem, only four transitions occur: S1 →S2 , S2 →S3 , S3 →S2 and S3 →S4 , only P(S2 |S1 ), P(S3 |S2 ), P(S2 |S3 ), P(S4 |S3 ) are considered in the state transition diagram of Fig. 4. The state transition matrix is given as,

⎡

0.25x ⎢1+x

⎢ ⎢ 0.25x ⎢ 2+x FST = ⎢ ⎢ ⎢ 0.25x ⎢ ⎣2+x 0.25

1 + 0.25x 1+x

0.25x 1+x

0.25x 2+x

2 + 0.25x 2+x

1 + 0.25x 2+x

0.25x 2+x

0.25

0.25

0.25x 1+x

⎤

⎥ ⎥ ⎥ ⎥ ⎥ ⎥ 1 + 0.25x ⎥ ⎥ 2+x ⎦ 0.25x 2+x

(6)

0.25

If zn and zn +1 present two consecutive states, then zn+ 1 = zn FST . If S is a set containing the states of Femtolet i.e. {S1 , S2 , S3 , S4 }, after t time, the state is given as, t−1 t zn+t = zn+t−1 FST = (zn+t−2 FST )FST = (zn+t−3 FST )FST FST . . . = (zn+t−t FST )FST = zn FST

(7)

A. Mukherjee, D. De / Simulation Modelling Practice and Theory 62 (2016) 68–87

75

Fig. 5. UML sequence diagram of application offloading using Femtolet.

The probability of occurrence of a state after t time can be determined using Eq. (7). For example, if t = 1 and x = 3, then after t time the probability of occurrence of each state from the initial state is determined as,

⎡

0.75 ⎢ 4

FST = [1

0

0

⎢ ⎢ 0.75 ⎢ 5 0]⎢ ⎢ ⎢ 0.75 ⎢ ⎣ 5 0.25

⎤

1.75 4

0.75 4

0.75 4 ⎥

0.75 5

2.75 5

1.75 5

0.75 5

1.75 ⎥ ⎥ 5 ⎦

0.25

0.25

0.25

⎥

0.75 ⎥ ⎥ 5 ⎥ = [0.1875

⎥

0.4375

0.1875

0.1875]

(8)

From Eq. (8) it is observed that in this case the probability of occurrence of state 2 i.e. Active is high. In this way, the probability of transition to different states from a state can be obtained using the proposed model. 3.4. Application offloading using Femtolet The sequence diagram of offloading an application by a MD to a Femtolet is presented in Fig. 5. To pictorially depict the sequence diagram of application offloading, Unified Modeling Language (UML) [33] is used in this paper. As observed from Fig. 5, a user requests for an application through his or her MD. Then the MD requests for the application to the Femtolet. If the Femtolet has the ability to execute the application, it accepts the request and executes it. After the execution is over, the processing result is sent back to the MD and the user receives the result. If the Femtolet is unable to process the user’s request, it asks the cloud for processing the application. The cloud therefore executes the requested application and sends back the result to the Femtolet. After receiving the result from the cloud, the Femtolet forwards the result to the MD and the user gets it.

76

A. Mukherjee, D. De / Simulation Modelling Practice and Theory 62 (2016) 68–87 Table 1 Parameters used in power and latency calculation. Parameter

Definition

Ptmf Prmf Ptfcl Prfcl Ptclc Prclc Ptsfc Prsfc Ptfcc Prfcc Ptflc Prflc Dtmf Drmf Dtfcl Drfcl Dtclc Drclc Dtsfc Drsfc Dtfcc Drfcc Dtflc Drflc Ltmf Lrmf Ltfcl Lrfcl Ltclc Lrclc Ltsfc Lrsfc Ltfcc Lrfcc Ltflc Lrflc Pi Sc Sm Im I

Power required for unit amount of data transmission from a MD to the femtocell/Femtolet/SCceNB Power required for unit amount of data reception by the MD from the femtocell/Femtolet/SCceNB Power required for unit amount of data transmission from the femtocell to the cloudlet Power required for unit amount of data reception by the femtocell from the cloudlet Power required for unit amount of data transmission from the cloudlet to the cloud Power required for unit amount of data reception by the cloudlet from the cloud Power required for unit amount of data transmission from the SCceNB to the femtocloud Power required for unit amount of data reception by the SCceNB from the femtocloud Power required for unit amount of data transmission from the femtocloud to the cloud Power required for unit amount of data reception by the femtocloud from the cloud Power required for unit amount of data transmission from the Femtolet to the cloud Power required for unit amount of data reception by the Femtolet from the cloud Amount of data transmission from the MD to the femtocell/ Femtolet/ SCceNB Amount of data reception by the MD from the femtocell/Femtolet/ SCceNB Amount of data transmission from the femtocell to the cloudlet Amount of data reception by the femtocell from the cloudlet Amount of data transmission from the cloudlet to the cloud Amount of data reception by the cloudlet from the cloud Amount of data transmission from the SCceNB to the femtocloud Amount of data reception by the SCceNB from the femtocloud Amount of data transmission from the femtocloud to the cloud Amount of data reception by the femtocloud from the cloud Amount of data transmission from the Femtolet to the cloud Amount of data reception by the Femtolet from the cloud Latency in unit amount of data transmission from the MD to the femtocell/Femtolet/SCceNB Latency in unit amount of data reception by the MD from the femtocell/Femtolet/SCceNB Latency in unit amount of data transmission from the femtocell to the cloudlet Latency in unit amount of data reception by the femtocell from the cloudlet Latency in unit amount of data transmission from the cloudlet to the cloud Latency in unit amount of data reception by the cloudlet from the cloud Latency in unit amount of data transmission from the SCceNB to the femtocloud Latency in unit amount of data reception by the SCceNB from the femtocloud Latency in unit amount of data transmission from the femtocloud to the cloud Latency in unit amount of data reception by the femtocloud from the cloud Latency in unit amount of data transmission from the Femtolet to the cloud Latency in unit amount of data reception by the Femtolet from the cloud Power consumption per unit time in executing instruction Speed of the cloud/cloudlet/Femtolet/SCceNB/femtocloud Speed of the MD Number of instructions executed inside the MD Number of instructions executed outside the MD

4. Power and latency consumption in Femtolet The parameters used in power and latency calculation are defined in Table 1.

4.1. Power consumption in Femtolet scenario The power consumption while the MD is communicating with the cloud through the Femtolet, is given as,

Pf l1 = (Ptm f × Dtm f + Prm f × Drm f ) + (Pt f lc × Dt f lc + Pr f lc × Dr f lc )

(9)

The power consumption in computation is given as,

I  m

Pf l2 = Pi

Sm

+

Pi

Ifl Ic + Pi Sc Sc

(10)

where Ifl and Ic are the number of instructions executed inside the Femtolet and cloud respectively and I = Ifl + Ic . Therefore the total power consumption considering the communication and computation is given by,

Pf l = Pf l1 + Pf l2

(11)

A. Mukherjee, D. De / Simulation Modelling Practice and Theory 62 (2016) 68–87

77

4.2. Latency in Femtolet scenario The latency involved while the MD is accessing the cloud services through the Femtolet, is calculated as,



L f l = (Ltm f × Dtm f + Lrm f × Drm f ) + (Lt f lc × Dt f lc + Lr f lc × Dr f lc ) +

Ifl Im Ic + + Sm Sc Sc

(12)

4.3. Power consumption in Femtocell plus cloudlet scenario The power consumption while the MD is communicating with the cloud through the femtocell and cloudlet, is calculated as,

Pf cl1 = (Ptm f × Dtm f + Prm f × Drm f ) + (Pt f cl × Dt f cl + Pr f cl × Dr f cl ) + (Ptclc × Dtclc + Prclc × Drclc )

(13)

The power consumption in computation is given as,

I  m

Pf cl2 = Pi

Sm

I cl

+ Pi

Sc

+ Pi

Ic Sc



(14)

where Icl and Ic are the number of instructions executed inside the cloudlet and cloud respectively and I = Icl + Ic . Therefore the total power consumption considering the communication and computation is given by,

Pf cl = Pf cl1 + Pf cl2

(15)

4.4. Latency in Femtocell plus cloudlet scenario The latency involved while the MD is accessing the cloud services through the femtocell and cloudlet, is calculated as,

L f cl = (Ltm f × Dtm f + Lrm f × Drm f ) + (Lt f cl × Dt f cl + Lr f cl × Dr f cl ) + (Ltclc × Dtclc + Lrclc × Drclc ) +

I

m

Sm

+

Icl Ic + Sc Sc



(16) 4.5. Power consumption in SCceNB scenario The power consumption while the MD is communicating with the cloud through the SCceNB and femtocloud, is calculated as,

Psc1 = (Ptm f × Dtm f + Prm f × Drm f ) + (Pts f c × Dts f c + Prs f c × Drs f c ) +(Pt f cc × Dt f cc + Pr f cc × Dr f cc )

(17)

The power consumption in computation is given as,

I  m

Psc2 = Pi

Sm



+

Ifc Is Ic Pi + Pi + Pi Sc Sc Sc

(18)

where Is , Ifc and Ic are the number of instructions executed inside the SCceNB, femtocloud and cloud respectively and I = Is + Ifc + Ic . Therefore the total power consumption considering the communication and computation is given by,

Psc = Psc1 + Psc2

(19)

4.6. Latency in SCceNB Scenario The latency involved while the MD is accessing the cloud services through the SCceNB and femtocloud, is given as,

Lsc = (Ltm f × Dtm f + Lrm f × Drm f ) + (Lts f c × Dts f c + Lrs f c × Drs f c )



+ (Lt f cc × Dt f cc + Lr f cc × Dr f cc ) +

Ifc Im Is Ic + + + Sm Sc Sc Sc

(20)

4.7. Comparison of power consumption and latency between Femtolet, femtocell plus cloudlet and SCceNB scenarios In (femtocell+cloudlet) scenario, computation offloading takes place either inside the cloudlet or inside the cloud as shown in Fig. 6. Comparing Eqs. (11) and (15) it is observed that Pfcl > Pfl . Similarly comparing Eqs. (12) and (16) it is observed that Lfcl > Lfl . In SCceNB, the base station provides limited computation. Therefore in SCceNB offloading can take place inside the base station itself. But the probability of offloading to the SCceNB is low because it has limited cloud functionalities. In that case offloading occurs inside the femtocloud or to the cloud as shown in Fig. 6. But Femtolet itself can handle

78

A. Mukherjee, D. De / Simulation Modelling Practice and Theory 62 (2016) 68–87

Fig. 6. Block diagram of (femtocell [10,18] + cloudlet [13,14]), SCceNB [15–17] and Femtolet scenarios.

Fig. 7. Power consumption vs. data size.

Fig. 8. Latency vs. data size.

A. Mukherjee, D. De / Simulation Modelling Practice and Theory 62 (2016) 68–87

79

Table 2 Comparison between proposed Femtolet and existing (Femtocell+cloudlet) and SCceNB scenarios. Feature Scenario

Advantages and disadvantages

Femtocell [10,18] + Cloudlet [13,14] • Femtocell is connected with cloudlet • Cloudlet is connected with cloud • Mobile device registered under the femtocell accesses the cloud services using cloudlet.

Advantages: • Femtocell provides good signal strength • Cloudlet provides high speed cloud service Disadvantages: • Communication delay between femtocell and cloudlet • Communication power between femtocell and cloudlet

SCceNB [15–17] • SCceNB is a home node base station with limited cloud functionalities • SCceNB is connected with femtocloud having intermediate storage and computation ability • Femtocloud is connected with cloud • Mobile device registered under the SCceNB accesses the cloud services using SCceNB and femtocloud. Advantages: • SCceNB provides good signal strength and limited cloud service • Femtocloud provides high speed cloud service • SCceNB provides communication and limited computation facilities simultaneously Disadvantages: • Due to limited cloud functionalities of SCceNB and intermediate storage and computation ability of femtocloud, sometime cloud is asked to provide service to the user. Access to remote cloud increases delay and power consumption. • Communication delay between SCceNB and femtocloud. • Communication power between SCceNB and femtocloud.

Power consumption Latency Remarks

Proposed Device Femtolet • Femtolet is a home node base station with storage and computation ability like cloudlet • Femtolet is connected with cloud • Femtolet integrates the functionalities of femtocell and cloudlet • Mobile device registered under the Femtolet accesses the cloud services using Femtolet.

Advantages: • Femtolet provides good signal strength and high speed cloud service at the same time as it contains the features of both femtocell and cloudlet. • Femtolet provides communication and computation facilities simultaneously. • As Femtolet contains the cache copies of data and code available inside the cloud, most of the time Femtolet itself serves the user. The probability of cloud access to serve the user request is very less. As a result the delay and power consumption are very less in case of Femtolet.

0.0198–0.1979 W approximately in 0.0185–0.1846 W approximately in 0.0165–0.1646 W approximately in accessing 1–10 TB data accessing 1–10 TB data accessing 1–10 TB data 0.0153–0.1533 s approximately in 0.0133–0.1333 s approximately in 0.0167-0.1667 s approximately in accessing 1–10 TB data accessing 1–10 TB data accessing 1–10 TB data • Femtolet reduces the latency by approximately 20% and 13% respectively than existing (femtocell+cloudlet) and SCceNB scenarios. • Femtolet reduces the power consumption by approximately 17% and 11% respectively than existing (femtocell+cloudlet) and SCceNB scenarios.

both of the communication and computation purposes. The Femtolet asks the cloud for computation only if it is unable to fulfill the user’s request. As the Femtolet contains cache copies of data and code stored inside the cloud, the probability of offloading to the cloud is very low. Therefore (Dtsfc + Dtfcc ) > Dtflc and (Drsfc + Drfcc ) > Drflc . Thus it can be predicted that Femtolet consumes low power and low latency than SCceNB. Comparing Eqs. (11) and (19), it can be predicted that Psc >Pfl . Comparing Eqs. (12) and (20), it can be predicted that Lsc > Lfl . Hence it is proved that Femtolet can provide cloud services with communication facility at low power and low latency than the existing (femtocell+cloudlet) and SCceNB scenarios. Fig. 7 presents the power consumption in case of Femtolet, (femtocell+cloudlet) and SCceNB scenarios, determined using Eqs. (11), (15) and (19) respectively. Fig. 8 draws a comparison between the latencies involved in Femtolet, (femtocell+cloudlet) and SCceNB scenarios, determined using Eqs. (12), (16) and (20) respectively. The power consumption is measured in watt (W) and latency is measured in second (s). The data size is measured in terabyte (TB). It is observed from Fig. 7 that using Femtolet the power consumption can be reduced by approximately 17% and 11% than existing (femtocell+cloudlet) and SCceNB architectures respectively. Fig. 8 demonstrates that Femtolet can reduce the latency by

80

A. Mukherjee, D. De / Simulation Modelling Practice and Theory 62 (2016) 68–87 Table 3 Parameters used in simulation. Layer

Parameter

Value

Physical layer

Radio type Packet reception model Antenna model Temperature Noise Factor MAC protocol Network protocol Routing protocol Sending and receiving buffer size (bytes) Battery model Battery change monitoring interval Full battery capacity (mAh) Simulation time (s) Item size(bytes)

802.11b radio PHY 802.11b reception model Omni directional 290.0 K 10.0 802.11 IPV4 Bellman Ford 512,8162,16384 (for MD, Femtolet and cloud respectively) Linear model 60 s 1200 300 512

MAC layer Network layer Transport layer Battery model

Scenario properties CBR properties

approximately 20% and 13% than existing (femtocell+cloudlet) and SCceNB architectures respectively. Therefore it is observed that Femtolet is a low power i.e. green and fast device for 5G mobile network. Table 2 presents a comparative analysis between the proposed Femtolet and existing SCceNB and (femtocell+cloudlet) scenarios. Hence it is observed that Femtolet reduces both of the power consumption and latency than the existing (femtocell+cloudlet) and SCceNB scarios. Therefore Femtolet is a green and fast device. 5. Performance analysis of Femtolet using Network Simulator Qualnet version 7 The simulation parameters used to simulate the working model of Femtolet in Qualnet version 7 are defined in Table 3. In this simulation model, the MD sends request for an application execution to the Femtolet i.e. base station with storage and computation facility. If the Femtolet is able to execute the application, it sends the processing result back to the MD. But if it is unable to provide solution to the user’s request, it sends the request to the cloud which provides the required service. Most of the time the Femtolet provides the services required by the user without intervention the cloud offering quick response. The Femtolet scenario is presented in Fig. 9. In this scenario, the mobile device i.e. MD is sending request to the Femtolet which transmits 10–100 mW power. If the Femtolet is unable to satisfy the user’s demand, the request is forwarded to the cloud. 5.1. Average delay The delay of a network is defined as the time taken by the data to travel from the sender to the receiver. Table 4 and Fig. 10 present the average delay of the proposed device Femtolet and cloud. The packet size is taken as 512 bytes. As observed from Table 4 and Fig. 10 that using Femtolet the average delay can be reduced by 35% approximately with respect to the cloud’s one. As a result the quality of user experience can be improved.

Fig. 9. Simulation scenario of Femtolet using Qualnet7.

A. Mukherjee, D. De / Simulation Modelling Practice and Theory 62 (2016) 68–87

81

Table 4 Average delay of Femtolet and cloud. Number of request packet sent

50 100 150 200 250

Average delay (s)

Mean delay (s)

Femtolet

Cloud

Femtolet

Cloud

0.0042 0.0045 0.0046 0.0057 0.0069

0.0075 0.008 0.008 0.0085 0.0088

0.00518

0.00816

Fig. 10. Average delay of Femtolet and cloud.

5.2. Average jitter Jitter is the delay in transmission of the current packet minus the delay in transmission of the previous packet. Average jitter is the division of the total received packet jitter by the total number of packets received minus 1. Table 5 and Fig. 11 present the average jitter for the proposed device Femtolet and cloud. The packet size is taken as 512 bytes. As observed from Table 5 and Fig. 11 that using Femtolet the average jitter can be reduced by 60% approximately than the cloud. Hence the quality of user experience can be improved. 5.3. Energy consumption In our simulation the energy consumptions of Femtolet in transmit and receive modes are determined. Table 6 and Fig. 12 present the energy consumed by the proposed device Femtolet in transmit and receive modes. The packet size is taken as 512 bytes. 5.4. Unicast received throughput and carried load The average rate of successful message delivery over a network is called throughput. It is measured in bits/s. Carried load is defined as the load carried by each node to transmit data in the network. It is measured in bits/s. Table 7 and Fig. 13 present the unicast received throughput and the carried load by the Femtolet. The packet size is taken as 512 bytes.

Table 5 Average jitter of Femtolet and cloud. Number of request packet sent

50 100 150 200 250

Average Jitter (s)

Mean jitter (s)

Femtolet

Cloud

Femtolet

Cloud

0.0024 0.0026 0.0027 0.0035 0.0038

0.0064 0.0073 0.0075 0.0082 0.0082

0.003

0.00752

82

A. Mukherjee, D. De / Simulation Modelling Practice and Theory 62 (2016) 68–87

Fig. 11. Average jitter of Femtolet and cloud.

Table 6 Energy consumption of Femtolet. Number of request packet sent

50 100 150 200 250

Energy consumption (mWh)

Mean energy consumption (mWh)

Transmit mode

Receive mode

Transmit mode

Receive mode

0.0121 0.0136 0.0158 0.0173 0.0182

0.0 0 01 0.0 0 03 0.0 0 06 0.0 0 08 0.001

0.0154

0.0 0 056

Table 7 Unicast Received throughput and carried load of Femtolet. Number of request packet sent

Unicast received throughput (bits/s)

Carried load (bits/s)

Mean unicast received throughput (bits/s)

50 100 150 200 250

4115 4865 6200 7255 8200

995 1510 2195 3205 3895

6127

Fig. 12. Energy consumption of Femtolet in transmit and receive modes.

Mean carried load (bits/s) 2360

A. Mukherjee, D. De / Simulation Modelling Practice and Theory 62 (2016) 68–87

83

Fig. 13. Unicast received throughput and carried load of Femtolet. Table 8 Offloading time and power for matrix inversion using Femtolet. Matrix order

Offloading time (s)

Offloading power (W)

Mean offloading time (s)

Mean offloading power (W)

2×2 3×3 4×4 5×5 6×6

4.98 10.15 14.75 23.04 30.01

0.22 0.63 0.79 1.18 1.45

16.586

0.854

Fig. 14. Offloading time for matrix inversion using Femtolet.

5.5. Offloading time and offloading power In our analysis the task of inverting matrix of different order is offloaded to the Femtolet. The time and power consumption in offloading this task using Femtolet are collected and presented in Table 8. Figs. 14 and 15 present the offloading time and power for matrix inversion using Femtolet. 5.6. Summary of simulation results Table 9 presents the summary of the simulation results shown in Figs. 10–15. As observed from Table 9, the average delay and average jitter are approximately 0.0 042–0.0 069 and 0.0 024–0.0 038 respectively for Femtolet. Table 9 presents that the energy consumed by Femtolet in transmit and receive modes are approximately 0.0121–0.0182 and 0.0 0 01–0.0 01 respectively. In the transmit mode, the Femtolet serves as a base station with cloud environment. During receive mode, it receives requests from the MDs and services from the cloud. As in transmit mode the Femtolet provides the communication and computation facilities to the MDs continuously, the energy consumption is more in the transmit mode than the receive mode. Table 9 shows that the unicast received throughput and carried load for Femtolet are approximately

84

A. Mukherjee, D. De / Simulation Modelling Practice and Theory 62 (2016) 68–87

Fig. 15. Offloading power for matrix inversion using Femtolet. Table 9 Summary table based on simulation results. Parameter

Obtained value

Average delay (s) Average jitter (s) Energy consumed (mWh) Transmit mode Receive mode Unicast Received Throughput (bits/sec) Carried load (bits/sec) Offloading Time for matrix inversion (sec) Offloading Power for matrix inversion (W)

0.0 042–0.0 069 (for 50 to 250 number of request packets, each packet size is 512 bytes) 0.0 024–0.0 038 (for 50 to 250 number of request packets, each packet size is 512 bytes) 0.0121–0.0182 (for 50 to 250 number of request packets, each packet size is 512 bytes) 0.0 0 01–0.0 01 (for 50 to 250 number of request packets, each packet size is 512 bytes) 4115–8200 (for 50 to 250 number of request packets, each packet size is 512 bytes) 995–3895 (for 50 to 250 number of request packets, each packet size is 512 bytes) 4.98–30.01 (for 2 × 2 to 6 × 6 matrix) 0.22–1.45 (for 2 × 2 to 6 × 6 matrix)

4115-8200 bits/s and 995-3895 bits/s respectively. Table 9 presents that the time and power consumption in offloading the task of matrix inversion are approximately 4.98–30.01 s and 0.22–1.45 W respectively. 5.7. Comparison of proposed work with existing schemes The contributions of the proposed work with respect to the existing works on application offloading in mobile cloud network are presented in Table 10. From Table 10 it is observed that for the first time we have proposed a low power device that serves as a home base station integrated with high speed cloud services. This is the novelty of the paper which makes it different from the existing architectures for application offloading using mobile cloud network. 6. Future scope of Femtolet In this section the challenges for Femtolet are discussed as follows: Challenge 1: Protocol for Femtolet Femtocell uses HNBAP protocol. As Femtolet is itself a small, low power base station like femtocell integrated with cloud services, HNBAP can be used for registration and deregistration purposes. As Femtolet is connected to the cloud and is able to provide cloud services like cloudlet, IPV4 can be used. Therefore a new protocol is to be designed for Femtolet as a combination of HNBAP and IPV4. Challenge 2: Handoff between femtocell and Femtolet When a user registered under a femtocell/SCceNB/Femtolet moves into the coverage of another femtocell/SCceNB/Femtolet, then a handoff will occur. Global Positioning System (GPS) is vital for Femtolet as it enables location assistance for frequency assistance, timing for system clock functionality and location licensing ability. To facilitate the handoff between Femtolets or between Femtolet and femtocell, or between SCceNB and Femtolet, sniffer can be used as a simplified receiver. The process of handoff management for Femtolet based mobile network is also a challenging area. Challenge 3: Densification and interference management for Femtolet based 5G mobile network Femtolet is proposed to provide good signal strength and high speed cloud services at the same time at low power in order to enhance the quality of service as well as user experience. Femtolets are usually deployed inside the coverage area of a macrocell base station to offer good signal level and high speed cloud services at indoor region. But dense deployment of Femtolets increases radiation and causes interference that reduces the signal-to-interference-plus-noise ratio (SINR). This

A. Mukherjee, D. De / Simulation Modelling Practice and Theory 62 (2016) 68–87

85

Table 10 Comparison of proposed work with existing works on offloading in mobile cloud computing. Feature

Existing Works Mobile cloud computing based offloading [8,11,19–24]

Offloading using Femto-cloud network [9,27–29]

Offloading using cloudlet [13,14,25,26]

Offloading using femtocell [10,18] + cloudlet [13,14]

Offloading using SCceNB [15–17]

Scenario

MDs are connected to the cloud through cellular base station/ WiFi.

MDs are connected to the cloud through femtocell.

MDs are connected to the cloudlet.

MDs are connected to the cloudlet through femtocell.

Contribution

Applications are executed inside the cloud and the results are sent back to the MDs.

Applications are executed inside the cloud and the results are sent back to the MDs through the femtocell.

Applications are executed inside the cloudlet and the results are sent back to the MDs.

Applications are executed inside the cloudlet and the results are sent back to the MDs through the femtocell.

Limited/ Full storage and processing abilities are integrated with Base station Cloud services are incorporated into Base station fully Reduces power consumption of MD Reduces WAN delay Remarks

No

No

No

No

MDs are connected to the SCceNB which is a HNB with limited cloud functionalities. Applications are executed inside the SCceNB or a small cloud called femtocloud and the results are sent back to the MDs. Limited storage and processing abilities are integrated with Base station

Full storage and processing abilities are integrated with Base station

X

X

X

X

X















X

X









Offloading using proposed device Femtolet

MDs are connected to the Femtolet which is a HNB with cloud functionalities. Applications are executed inside the Femtolet and the results are sent back to the MDs.

The proposed device Femtolet has the full storage and processing abilities and this is not available in the existing frameworks. Femtolet is a base station which incorporates all the cloud services: IaaS, PaaS and SaaS. Therefore the novelty as well as contribution of our work is the device Femtolet which has been introduced for the first time in this paper for mobile network to provide communication and computation simultaneously at low power and low latency.

densification problem is a major research scope of 5G mobile network [1]. Fig. 16 shows the power transmission inside the coverage area of a macrocell containing Femtolets. It is observed that with the increase in number of allocated Femtolet, power transmission in the area is increased. The amount of power transmission in the area is determined by the sum of the transmission power of the macrocell base station and the power transmitted by all the allocated Femtolets. Fig. 16 shows the SINR for a Femtolet with respect to the number of its neighbour Femtolet. It is observed that with the increase in the number of adjacent Femtolet, the SINR degrades. Thus if large number of Femtolets are allocated in a small region, the SINR will be reduced. As a result the service quality will degrade. To deal with this problem, an energy-efficient and economic deployment of Femtolets in 5G mobile network is required that will maintain a trade-off between the interference and densification. Challenge 4: Recovery management for Femtolet When a Femtolet gets damaged, its user can be handed over to the neighbour Femtolets. If no Femtolet is available nearby, the user can search for a nearby cloudlet or directly offload to the cloud. This is also a research area where innovative and efficient approaches are required to develop. Challenge 5: Opportunistic selection for computation offloading in multi-Femtolet scenario The processing and storage capacity as well as current load of each Femtolet may vary. If a mobile device has to be connected with the network and has to offload an application and multiple Femtolets are available nearby, then the mobile device has to select the most suitable Femtolet providing better connectivity and the opportunity of offloading the application at minimum power and minimum latency than the other Femtolets. Therefore opportunistic selection in multi-Femtolet scenario for offloading an application is also a challenging area. Challenge 6: Collaboration with Millimeter wave, femotcell, and SCceNB in 5G mobile network The 5G mobile network deals with small cell base stations with short range communication. Femtocell and SCceNB both are small cell base station. Millimeter wave links are going to be used in small cell base station for high speed Internet connectivity [16]. Millimeter wave is used for high bandwidth short range communication [16,34]. As Femtolet itself is a small cell base station with cloud functionalities, collaboration is required between Femtolet and the existing technologies.

86

A. Mukherjee, D. De / Simulation Modelling Practice and Theory 62 (2016) 68–87

Fig. 16. Effect of dense deployment of Femtolet inside the macrocell coverage area.

Challenge 7: Cooperation with multiple cells in 5G mobile network As 5G mobile network contains different types of base stations, cooperation between them is essential [16]. When offloading takes place using Femtolet, both communication and computational resources are involved. In such situation, cooperation between the cells is required for reducing latency as well as reducing interference. The Femtolets, femtocells and SCceNBs should work in a self-organized way so that the network becomes stable [35]. In cooperative femtocell network [35], the femtocells cooperate with each other in such a way that the system utility is maximized. If Femtolets are used along with femtocells and SCceNBs, how the cooperation among them will take place to maximize the system utility is another challenging issue. 7. Conclusion To provide high speed and low power Internet access in 5G mobile network, this paper has introduced a new network device denoted as Femtolet. The architecture and working principle of Femtolet have been proposed. The working model of Femtolet is simulated in network simulator Qualnet version 7 and its performance is analyzed with respect to average delay, jitter, energy consumption, carried load, throughput, offloading time and power. Femtolet is a home base station that contains large internal storage with computation ability. Femtolet provides high speed cloud services to the user like cloudlet as well as provides high signal level like femtocell base station. The mobile users registered under a Femtolet can make and receive call, send messages as well as can offload data and applications to the Femtolet. From the mathematical analysis based on the assumed parameter values, it is observed that using Femtolet instead of existing femtocell plus cloudlet and SCceNB scenarios, power consumption can be reduced by approximately 17% and 11% respectively. It is also demonstrated

A. Mukherjee, D. De / Simulation Modelling Practice and Theory 62 (2016) 68–87

87

that Femtolet can reduce the latency by approximately 20% and 13% respectively than existing femtocell plus cloudlet and SCceNB scenarios. Therefore Femtolet can be referred as a 5G network device that will provide fast and green mobile cloud computing. Acknowledgment Authors are grateful to Department of Science and Technology (DST) for sanctioning a research Project entitled “Dynamic Optimization of Green Mobile Networks: Algorithm, Architecture and Applications” under Fast Track Young Scientist scheme reference no.: SERB/F/5044/2012-2013, DST-FIST reference no.: SR/FST/ETI-296/2011, and No.DST/INSPIRE Fellowship/2013/327 under which this article is completed. References [1] H.S. Dhillon, R.K. Ganti, F. Baccelli, J.G. Andrews, Modeling and analysis of K-tier downlink heterogeneous cellular networks, IEEE J. Sel. Areas Commun. 30 (2012) 550–560. [2] W.H. Chin, Z. Fan, R. Haines, Emerging technologies and research challenges for 5G wireless networks, IEEE Wirel. Commun. 21 (2014) 106–112. [3] M.N. Tehrani, M. Uysal, H. Yanikomeroglu, Device-to-device communication in 5G cellular networks: challenges, solutions, and future directions, IEEE Commun. Mag. 52 (2014) 86–92. [4] A. Mukherjee, S. Bhattacherjee, S. Pal, D. De, Femtocell based green power consumption methods for mobile network, Comput. Netw. 57 (2013) 162– 178. [5] A. Mukherjee, D. De, Congestion detection, prevention and avoidance strategies for an intelligent, energy and spectrum efficient green mobile network, J. Comput. Intell. Electron. Syst. 2 (2013) 1–19. [6] Q. Zhang, Z. Feng, W. Li, Coverage self-optimization for randomly deployed femtocell networks, Wirel. Person. Commun. 82 (2015) 2481–2504. [7] A. Mukherjee, D. De, A novel cost-effective and high-speed location tracking scheme for overlay macrocell-femtocell network, in: Proceedings of the URSIGA, IEEE, 2014, pp. 1–4. [8] N. Fernando, S.W. Loke, W. Rahayu, Mobile Cloud Computing: A Survey, Future Gener. Comput. Syst. 29 (2013) 84–106. [9] D. De, A. Mukherjee, Femto-cloud based secure and economic distributed diagnosis and home health care system, J. Med. Imag. Health Inf. 5 (2015) 433–447. [10] A. Mukherjee, P. Gupta, D. De, Mobile cloud computing based energy efficient offloading strategies for femtocell network, in: Proceedings of the Applications and Innovations in Mobile Computing, IEEE, 2014, pp. 28–35. [11] H.W. Lv, J.Y. Lin, H.Q. Wang, G.S. Feng, M. Zhou, Analyzing the service availability of mobile cloud computing systems by fluid-flow approximation, Front. Inf. Technol. Electron. Eng. 16 (2015) 553–567. [12] H. Qi, M. Shiraz, J.Y. Liu, A. Gani, Z.A. Rahman, T.A. Altameem, Data center network architecture in cloud computing: review, taxonomy, and open research issues, J. Zhejiang Univ. Sci. C 15 (2014) 776–793. [13] M. Satyanarayanan, P. Bahl, R. Caceres, N. Davies, The case for VM-Based cloudlets in mobile computing, IEEE Pervasive Comput. 8 (2009) 14–23. [14] T. Verbelen, S. Pieter, D.T. Filip, D. Bart, Cloudlets: bringing the cloud to the mobile user, in: Proceedings of the Third ACM Workshop on Mobile Cloud Computing and Services, ACM, 2012, pp. 29–36. [15] O.M. Murioz, A. Pascual-Iserte, J. Vidal, Joint allocation of radio and computational resources in wireless application offloading, in: Proceedings of the Future Network and Mobile Summit, IEEE, 2013, pp. 1–10. [16] S. Barbarossa, S. Sardellitti, P. Di Lorenzo, Communicating while computing: distributed mobile cloud computing over 5G heterogeneous networks, IEEE Sign. Process. Mag. 31 (2014) 45–55. [17] F. Lobillo, Z. Becvar, M.A. Puente, P. Mach, F. Lo Presti, F. Gambetti, M. Goldhamer, J. Vidal, A.K. Widiawan, E. Calvanesse, An architecture for mobile computation offloading on cloud-enabled LTE small cells, in: Proceedings of the Wireless Communications and Networking Conference Workshops, IEEE, 2014, pp. 1–6. [18] I. Ashraf, T.W. Ho, H. Claussen, Improving energy efficiency of femtocell base stations via user activity detection, in: Proceedings of the WCNC, IEEE Communications Society, 2010, pp. 1–5. [19] H.T. Dinh, C. Lee, D. Niyato, P. Wang, A survey of mobile cloud computing: architecture, applications, and approaches, Wirel. Commun. Mobile Comput. 13 (2013) 1587–1611. [20] S. Abolfazli, Z. Sanaei, E. Ahmed, A. Gani, R. Buyya, Cloud-based augmentation for mobile devices: motivation, taxonomies, and open challenges, IEEE Commun. Surv. Tutor. 16 (2014) 337–368. [21] E. Ahmed, A. Gani, M.K. Khan, R. Buyya, S.U. Khan, Seamless application execution in mobile cloud computing: motivation, taxonomy, and open challenges, J. Netw. Comput. Appl. 52 (2015) 154–172. [22] E. Ahmed, A. Akhunzada, M. Whaiduzzaman, A. Gani, S.H. Ab Hamid, R. Buyya, Network-centric performance analysis of runtime application migration in mobile cloud computing, Simul. Model. Pract. Theory 50 (2015) 42–56. [23] M. Amoretti, A. Grazioli, F. Zanichelli, A modeling and simulation framework for mobile cloud computing, Simul. Model. Pract. Theory 58 (2015) 140–156. [24] N. Aminzadeh, Z. Sanaei, S.H. Ab Hamid, Mobile storage augmentation in mobile cloud computing: taxonomy, approaches, and open issues, Simul. Model. Pract. Theory 50 (2015) 96–108. [25] S. Bohez, T. Verbelen, P. Simoens, B. Dhoedt, Discrete-event simulation for efficient and stable resource allocation in collaborative mobile cloudlets, Simul. Model. Pract. Theory 50 (2015) 109–129. [26] F.R. Duro, J.G. Blas, D. Higuero, O. Perez, J. Carretero, CoSMiC: A hierarchical cloudlet-based storage architecture for mobile clouds, Simul. Model. Pract. Theory 50 (2015) 3–19. [27] S. Bhattacherjee, S. Majumder, D. De, Trust model for femto-cloud based mobile network, in: Proceedings of the Fifth International Conference on The Next Generation Information Technology Summit, IEEE, 2014, pp. 53–58. [28] A. Mukherjee, D. De, Low power offloading strategy for femto-cloud mobile network, Eng. Sci. Technol., Int. J. (2015), doi:10.1016/j.jestch.2015.08.001. [29] D. De, A. Mukherjee, S. Bhattacherjee, P. Gupta, Trusted cloud-and femtocell-based biometric authentication for mobile networks, in: Handbook of Research on Securing Cloud-Based Databases with Biometric Applications, IGI Global, 2014, p. 320. [30] A. Mukherjee, D. De, P. Deb, Interference management in macro-femtocell and micro-femtocell cluster based LTE-advanced green mobile network, IET Commun. (2016), doi:10.1049/iet-com.2015.0982. [31] J. Han, M. Kamber, Data Mining Concepts and Techniques, Morgan Kaufmann Publishers, Second Edition, Elsevier, 2010. [32] S. Karlin, A First Course in Stochastic Processes, Academic Press, 2014. [33] J. Rumbaugh, I. Jacobson, G. Booch, Unified Modeling Language Reference Manual, The Pearson Higher Education, 2014. [34] T.S. Rappaport, S. Sun, R. Mayzus, H. Zhao, Y. Azar, K. Wang, G.N. Wong, J.K. Schulz, M. Samimi, F. Gutierrez, Millimeter wave mobile communications for 5G cellular: it will work!, IEEE Access 1 (2013) 335–349. [35] F. Pantisano, M. Bennis, W. Saad, M. Debbah, M. Latva-aho, Interference alignment for cooperative femtocell networks: a game-theoretic approach, IEEE Trans. Mobile Comput. 12 (2013) 2233–2246.