Академический Документы
Профессиональный Документы
Культура Документы
hotspot coverage for capacity enhancement. The cations in HetNet reliably. We first introduce a
reduced coverage area of small cells means that framework of HetNet needed to fulfill MCC ser-
the number of MDs sharing the same cell is vice requirements. Then, several challenges are
lower compared to macro cells, giving MDs discussed for HetNet in supporting MCC appli-
more freedom of using the bandwidth with lower cations. We also envision the future for MCC in
transmission power. HetNet before drawing the conclusion.
In this article, we consider smart mobile
devices with handheld size and limited comput- FRAMEWORK OF HETNET FOR MCC
ing power, which are efficiently connected to the
Internet by HetNet. The full potential of mobile In MCC, MDs connect to the Internet and then
cloud computing applications can be explored to the cloud via the wireless connections. There-
only when computation and storage are offload- fore, the wireless network is an important com-
ed into the cloud with acceptable latency and ponent of the MCC architecture. With SMDs
overhead, and doesn’t disturb user interactivity becoming increasingly ubiquitous and mobile
with the mobile applications. As the wireless application economy continuing to show impres-
environments may change, the application has to sive growth, future wireless networks have to
shift its computation workload between MD and support exploding mobile data traffic and should
cloud without operation interruptions, consider- be optimized for mobile broadband traffic. The
ing the time-varying wireless connections in Het- introduction of MCC puts further strain on wire-
Net. To deliver cloud services in HetNet less networks since additional data associated
environment, we have to face several problems. with cloud services between the MDs and the
For example, wireless networks are not always cloud have to be transported via the wireless
reliable to guarantee cloud service delivery due networks with potential QoS and/or power con-
to straints.
• User mobility HetNets offer promising solutions for these
• Propagation effects of wireless channels challenges. In this section, we discuss the frame-
• Traffic load variations affecting both the multi- work of HetNet for MCC, identifying the main
user resource sharing within a serving cell and functional blocks. We especially focus on the tra-
interference across neighboring cells ditional centralized cloud, of which the comput-
So, MDs have to collaborate with the eNodeBs ing resource pool is placed in the remote cloud
in making offloading decisions. Also, the and MDs can access the resources by connecting
eNodeBs have to guarantee the QoS of cloud to the existing wireless network. As illustrated by
service transmission with proper radio resource Fig. 1, the HetNet mainly consists of two compo-
management (RRM) decisions such as admis- nents, i.e., macrocells and small cells, where the
sion control, cell association, power control and former provide mobility while the latter boost
resource allocation. To the best of the authors’ coverage and capacity.
knowledge, there have been quite few work in
the existing literatures on these problems in Het- ACCESS NODES IN HETNET
Net for MCC applications. Macro/Micro Cells — The inter-site distance (ISD)
The scope of this article is hence to examine between two macro or micro eNBs (MeNBs) is
how SMDs can work so as to enjoy MCC appli- usually no less than 500 meters, which can pro-
In order to enable vide a ubiquitous coverage for MDs in a wide network. For example, a task with high computa-
range of area, and support high mobility MDs by tion and low communication requirements is
applications and systems minimizing the handover frequency. However, more likely to benefit from offloading than a
due to factors such as channel fading and traffic task with low computation and high communica-
to continue to operate in congestion, the connectivity between MDs and tion requirements. Therefore, a judicious deci-
dynamic wireless MeNB has relatively low data rate and is unsta- sion has to be made on whether to offload a
ble. computation task or not.
environments, mobile In order to enable applications and systems
cloud applications must Small Cells — Since SMDs and cloud applications to continue to operate in dynamic wireless envi-
are hungry for a high-speed and stable connec- ronments, mobile cloud applications must
dynamically adjust the tivity to the cloud, the small cells become a bet- dynamically adjust the computing functionalities
ter choice. between the MD and cloud depending on the
computing functionalities changes in mobile environments. However, it is
between the MD and Pico Cells — The low-powered radio access hard to predict the data transmission delay of
nodes, which have a coverage range of about 200 wireless network and the power consumption of
cloud depending on the meters or less, are deployed in Pico cells. Their MD for computation offloading when an offload-
changes in mobile access is open to all cellular MDs. Usually, the ing decision has to be made. Existing dynamic
pico cells are used to provide hotspot coverage offloading approaches mostly use the measured
environments. in malls, airports or stadiums. metrics before and at the decision instant as a
reference and assume that the network condition
Femto Cells — A femto cell has a small, low- will remain constant during the execution inter-
power base station, typically designed for MDs val, which is apparently inaccurate [6, 7]. Anoth-
in a home or small business, whose access node er method is to model the wireless channel as a
is referred to as Home eNB in LTE. Similar to Markov chain to capture its fading effects [8, 9].
WiFi, the coverage range for a femto cell is less However, since the wireless resources are usually
than 100 meters. The access to a femto cell can shared by multiple MDs in HetNet, the data
be either restrained to a limited set of MDs in transmission delay and power consumption of a
the femto cell’s access control list (closed access MCC application are also dependent on the traf-
mode), or open to all cellular MDs (open access fic load and interference conditions of the asso-
mode). ciated cell, which are neglected in the above
studies.
Distributed Antenna System — A distributed- In this article, we propose a framework of
antenna system is a network of spatially separat- HetNet for MCC as shown in Fig. 2, where the
ed antenna nodes connected to a common source HetNet offers to the offloading decision function
via a transport medium that provides wireless a range of wireless transmission services with dif-
service within a geographic area or structure. It ferent serivce classes and monetary prices, and
can create small virtual cells by distributing the offloading decision is made considering both
antennas of macro eNBs across entire cell. The the offloading gain and the cost of using the Het-
antennas are connected to a common processing Net when a Service Level Agreement (SLA) is
unit via fiber. established with it. Specifically, the SLAs are
based on the estimated average values or worst
Relay Nodes — The relay nodes are low power case values with a given violation probability of
base stations that can provide coverage and two attributes, i.e., power consumption and wire-
capacity enhancement to macro cells at the cell less transmission delay. The former is the total
edge. A relay node is connected to its Donor power consumed in the MD and the latter is the
eNB (DeNB) via a radio interface. Due to the total delay via the HetNet for transmitting the
instability of DeNB coverage, MDs in some loca- computation tasks to the cloud and receiving the
tions of the macro cell may have a failure in computation results from the cloud. Although
access to the DeNB. The deployment of relay there are some existing research on the QoS
nodes can solve this problem effectively. framework for MCC [4, 5], we focus on the inter-
actions between the offloading decision function
MAIN FUNCTIONAL BLOCKS TO SUPPORT MCC of MCC and the radio resource management
Offloading Decision (OD) — With the continuous function of HetNet, which haven’t been addressed
enhancement of MDs’ capabilities, they gradual- in prior work to the best of our knowledge.
ly become the primary tools for accessing service Our general model for the SLA between the
clouds to maximize their functionalities. The two OD function and the HetNet consists of a tuple
most widely studied cloud services are cloud L i, L r, P, T, C, x p, x t, where L i and L r are the
storage services and cloud processing services. maximum amount of allowed data for computa-
Computation offloading is an enabling technolo- tion tasks and results to be transmitted via the
gy for the resource limited SMDs to access the wireless uplink and downlink, respectively. P and
latter services, where processing capabilities of T are the average or worst case values of power
SMDs are augmented by outsourcing computa- consumption and wireless transmission delay,
tion intensive components of the mobile applica- respectively. C is the service cost for data trans-
tions to the resourceful servers in clouds. The mission in HetNet, while xp and xt are the penal-
main purposes of computation offloading are to ty rates that HetNet will refund its users with for
save the SMDs’ power consumption and speed possible violations of the service class power
up the application processing. However, the consumption and wireless transmission delay,
above two goals may not always be achieved respectively. The OD function decides whether
depending on the particulars of the computation to offload a computation task and with what ser-
task, the server load, and the connectivity to the vice class. If the OD function makes a positive
No
Power control
resource allocation
No Meet the
New connection Penalty power and QoS
requirement?
Existing connection Yes
Remote
execution Reward
offloading decision, it sends the required service requirement may introduce interference to the
class to the HetNet, which decides whether to neighboring cells which reuse the same frequen-
accept the offloading requirement and associate cy with the serving cell, the existing connections
the SMD with a specific cell for data transmis- in both the serving cell and its neighboring cells
sion by the admission control and cell associa- should be taken into account. Finally, the
tion function. Once the HetNet accepts the reward, cost and SLA violation penalties should
offloading requirement and establishes the SLA all be considered in an optimized AC policy.
with the OD function, it provides QoS guarantee Since there are multiple types of access nodes
by the power control and resource allocation in HetNet, e.g., macro eNBs and pico eNBs, cell
function. association (CA) has to decide which access
node should be associated with the SMD for the
Admission Control (AC) and Cell Association (CA) — The offloading data transmission. The simplest policy
OD function in MCC requires that the HetNet is to associate the SMD with the access node
provides data transmission service with a power whose signal is received with the largest average
and QoS (e.g., delay) guarantee. However, as strength. Other more complex policies consider
the network traffic load varies with time, admis- the load balancing and interference between dif-
sion control is needed in HetNet to balance the ferent cells.
goals of maximizing bandwidth utilization and In our framework of HetNet for MCC, the
ensuring sufficient resources for the offloading AC and CA functions have to be performed
requirements with power and QoS constraints. jointly. When selecting the associated cell for a
The HetNet uses AC policies to determine the SMD with offloading requirement, the CA func-
admissibility of an offloading requirement from tion needs to consider whether this requirement
the OD function once it requests to establish a can be admitted by the AC function of a candi-
SLA for transferring the offloading data. The date cell, and an optimized cell is selected only
basic principle of the AC policy is to admit an among those cells considered eligible by the AC
offloading requirement only if the power and function.
QoS constraints of the new connection can be
met while the power and QoS guarantees for the Power Control (PC) and Resource Allocation (RA) —
existing connections will not be jeopardized. Once an offloading requirement is admitted into
Moreover, since admitting the new offloading the HetNet, power control and resource alloca-
Improved versions of a simple model [10] from • The impact of wireless network traffic load on the power
the following aspects: consumption and execution time for remote execution is
• Optimization objectives, ignored;
Offload Decision
• Energy and speed tradeoffs, • The cost charged and the penalty rates refunded by the
• Application partition, wireless network for providing and violating the QoS
• Dynamic variation of environment. guarantee are not considered.
Criteria for accepting/rejecting a connection: • The power consumption and wireless transmission delay
• Transmit power, should be considered as the criteria for acceptance/
• Network load, rejection/association decision;
• Achievable throughput, • Ensure that the required QoS be met when a MD roam to
Admission Control • QoS. a neighboring cell, especially during the download phase.
Design objectives:
• Minimization of both false rejections and false
admissions.
Design objectives:
• The interference introduced to the neighboring cells
• Best SNR,
should be considered.
• Load balance,
Cell Association • Throughput maximization under static
interference,
• Performance optimization under dynamic
interference.
tion functions are used to ensure its QoS guar- T corresponds to the sum of offloading SMD’s
antee whenever feasible. Resource allocation sojourn time in the queueing system for uplink
decides which MD is going to transmit on each transmission Tu and downlink transmission Td,
time-frequency resource in a cell, and power where the sojourn time of uplink or downlink
control determines the amount of power allocat- transmission includes both the waiting time Tuw
ed to each scheduled MD on the allocated for uplink or Tdw for downlink and service time
resources. The wireless transmission is per- Tus for uplink or Tds for downlink. On the other
formed on a slot-by-slot basis. In each time slot hand, the power consumption of an offloading
t, power control and resource allocation algo- requirement on a SMD corresponds to the sum
rithms normally make use of the current channel of power consumed during uplink and downlink
state S(t) and queue state Q(t) information when transmission, and also during the waiting period
making a decision. Given S(t), Q(t), and a power for remote cloud computation. Specifically, the
control and resource allocation action A(t) for power consumption for uplink transmission
uplink or downlink transmission, the correspond- includes both the power consumed for transmit-
ing uplink or downlink instantaneous transmis- ting computation task with power level P(t) dur-
sion rate R(t) and transmit power P(t) of each ing service time Tus and for receiving downlink
MD are determined, where R(t) can be seen as control signaling during waiting time Tuw. Mean-
its service rate in a queuing system. The wireless while, the power consumption for downlink
transmission delay of an offloading requirement transmission includes the power consumed for
receiving the computation results and downlink user interaction) and offloadable ones. Before a
control signaling during the downlink sojourn task is executed, it may require certain amount
time Td. Therefore, the power consumption and of data from other tasks. Therefore, data migra-
wireless transmission delay incurred by an tion via wireless networks is needed if this task
offloading requirement are determined by the and the other tasks are executed at different
power control and resource allocation algo- locations (i.e., MD or cloud server node).
rithms. For both offloading techniques, a controller
in the SMD needs to determine whether to
CHALLENGES IN HETNET FOR MCC offload the application or some of its computa-
tion tasks to the cloud when there is a request
In our proposed framework of HetNet for MCC, for application execution. The offloading deci-
every functional block needs to make a decision sion can be either made statically before execu-
based on the current system state to obtain the tion, or dynamically at runtime to cope with the
optimized performance. However, the current dynamic processing loads on SMD and cloud
state-of-art techniques for each functional block server nodes and the changing wireless channel
face several challenges in order to make the status and wireless network traffic load.
optimized decision, which are discussed in this In [10], a simple model is provided to analyze
section and summarized in Table 1. the energy saving achieved by computation
offloading. Suppose a computation task requires
OFFLOADING DECISION C instructions, define
As shown in Fig. 3, there are generally two kinds M as the speeds in instructions per second of
of offloading techniques, called as full offloading the MD,
and partial offloading. S as the speeds in instructions per second of
the cloud server,
Full Offloading — When full offloading, all compu- D as the bytes which the cloud server and MD
tation tasks of mobile applications are moved exchanged,
from the local MD to the remote cloud. This B as transmission rate of the MD via wireless
may significantly reduce the implementation network,
complexity of MDs, which makes the MDs Pc as the energy consumed by MD for comput-
lighter and smaller. However, different computa- ing,
tion tasks of a single application may have dif- Pi as the energy consumed by MD while being
ferent characteristics that make them more or idle,
less suitable for offloading. Therefore, full Ptr as the energy consumed by MD for sending
offloading is not always the optimal choice for and receiving data.
MDs. Then, the amount of energy saved is
The most important and of computation C are needed with relatively However, this is not true in partial offloading.
small amounts of communication D. However, As shown in Fig. 3, since the execution of com-
widely studied QoS this simple model is insufficient in providing an putation task 2 needs data input from computa-
optimized solution for offloading. Other existing tion task 0 and 1, the amount of data exchange
metric is the response work can be seen as improved versions of this D for the remote execution of task 2 depends on
delay or execution time simple model from various aspects, which are the execution locations of task 0 and task 1. In
discussed as follows. [6] and [7], a call graph and a profile tree are
of the application, constructed, respectively, to provide a global
where the remote Optimization Objectives — The simple model tries to view of the application behavior and specify the
get the maximum benefit by balancing the com- amount of data migration needed between dif-
execution time includes putation and communication power. However, ferent tasks. A dynamic partition algorithm
good QoS performance of the services has to be based on Lyapunov optimization is proposed in
transmission delay of guaranteed from the view of user experience. In [11] to make the offloading decision for all the
both the wireless other words, the computation-intensive tasks can computation tasks.
be offloaded to the cloud only when the QoS is
network and the ensured. The most important and widely studied Dynamic Variation of Environment — The simple
Internet, and task QoS metric is the response delay or execution model targets to find the optimal strategy by
time of the application, where the remote execu- using the transmission rate and execution time
execution delay of the tion time includes transmission delay of both the of computation statically. However, as the execu-
wireless network and the Internet, and task exe- tion time changes at different execution instances
remote cloud. cution delay of the remote cloud. Generally due to varying workload of the MD and cloud
speaking, there are five classes of optimization server nodes, it is hard to find an accurate static
objectives considering all possible combinations information. Moreover, the dynamic variation of
of these two metrics: the wireless channel condition along with the
1 Optimizing energy consumption, changing traffic load of the wireless network
2 Optimizing execution time, make the transmission rate highly dynamic. In
3 Optimizing energy consumption under execu- [11], dynamic offloading is studied considering
tion time constraint, the dynamic arrival of application execution
4 Optimizing execution time under energy con- requirement and the variation of data transmis-
sumption constraint, sion rate at different execution period. However,
5 Optimizing both energy consumption and exe- the data rate is assumed to remain constant
cution time. within a single application execution period. In
Objective 3) is addressed in [6, 7, 9, 11]. Objec- [8] and [9], the wireless channel variation within
tive 5) is a nontrivial multi-objective optimiza- the execution period of an application is mod-
tion problem. Reference [8] addressed objective eled by two-state Markov chains.
5) by setting the optimization objective as the Although existing work has made the above
total energy consumption plus the product of improvements to the simple model from various
total remote execution time and a predefined aspects, there are still some open problems,
delay/wait cost. By varying the delay cost, differ- especially when the offloading data is transmit-
ent power/energy versus delay trade-off balances ted via HetNet. First, the offloading decision is
can be obtained. made based on an estimation of the SMD’s
power consumption and execution time when
Energy and Speed Tradeoffs — In practice, the value the application is executed locally or remotely.
of M is dependent on the value of P c since low Existing work seldom considers the impact of
speed of the MD’s processor can provide signifi- wireless network traffic load on the above two
cant energy savings, while the value of B is parameters for remote execution. However in
dependent on the value of Ptr since the transmis- practice, the traffic load has an important influ-
sion rate of a wireless link is a function of its sig- ence over the remote execution time through its
nal-to-interference-noise-ratio (SINR) and impact on the waiting time of the queuing sys-
transmission bandwidth, where the SINR value tem corresponding to the wireless transmission.
is determined by the transmit power, channel On the other hand, the power consumption for
gain, and noise plus interference power. The remote execution may also be impacted by the
correlation between every pair of variables is not traffic load of the SMD’s serving cell and neigh-
explored in the simple model. In [8], the optimal boring cells. This is because the power consump-
values of P c and P tr are determined where it is tion is dependent on the channel gains of the
assumed that B = Ptr/(Ptr + i) with i represent- uplink wireless channel at the scheduling instants
ing the uplink channel stress. The closed-form of the SMD, which are likely to be affected by
solutions of two constrained dynamic optimiza- the traffic load. For example, the probability of
tion problems are derived in [9] to set the opti- the SMD being scheduled in a time slot with
mal clock frequency of local processor and data more favorable channel conditions becomes larg-
transmission rate of each time slot within the er with the increasing traffic load due to the
application execution interval, so that the power multiuser diversity effect if the channel-aware
consumption is minimized within a delay con- scheduling algorithms are applied. Moreover,
straint. The offload decision is made by choosing increased traffic load at the neighboring cells
the execution location with smaller energy con- may result in the larger interference opportuni-
sumption. ties for the SMD’s transmission, which in turn
leads to larger power consumption. We perform
Application Partition — In the simple model, the simulation on a HetNet consisting of one macro-
amount of data exchange between the server cell and three pico cells and also on a macro-
cloud and MD D is assumed to be a fixed value. only network, where users with computation
Through cooperation power eNodeB is small and thus the interfer- femto eNodeB. In this situation, the overall
ence situations may dynamically change, i.e. energy efficiency of the network becomes even
between different kinds some low power eNodeBs may not be active worse after deploying the femtocells. Interfer-
in a certain period of time since there are no ence management is therefore an important
of radio access nodes, data packets to transmit for the MDs associat- method to capitalize on the potential energy effi-
HetNet can perform ed with them. A cell association policy for ciency in HetNet. Various interference manage-
HetNet is proposed in [15] under the dynamic ment techniques, such as power control and
power control and interference assumption which aims to opti- interference coordination, can be used to pro-
resource allocation mize the average throughput or delay for the vide the promising energy-efficiency perfor-
MDs. mances.
of multiple cells When considering the support of MCC appli-
cations, several challenges exist for the AC and Power Control — Since the computation offloading
simultaneously to CA functions. First, in both the AC and CA func- decision is mainly concerned with the power
achieve maximum tions, power consumption and transmission delay consumption on the SMD for transmitting/
should be considered as the criteria for accep- receiving the offloading data to/from the cloud,
throughput and meet tance/rejection/association decision. Moreover, we will focus on the uplink power control mech-
the power and QoS since the transmission delay involves both the anisms as the power consumption for data recep-
computation task upload delay and computation tion is far less than that for data transmission.
requirements for results download delay, both the uplink and down- In 3GPP LTE, Fractional Path Loss Compen-
link traffic and wireless channel conditions have to sation Power Control (FPC) mechanism is used
MCC applications. be taken into account. Second, as a MD may roam for uplink power control, which is open-loop and
to a neighboring cell during the offloading data based on the path loss measurement done by the
transmission period, the AC function needs to MD but controlled with a factor a by the net-
ensure that the required QoS be met under this work. Meanwhile, a closed-loop power control
situation. In particular, as it is more intolerable for mechanism can also be applied, where measure-
an offloading task to be interrupted at the result ments by the eNodeBs are used to generate
download phase than at the task upload phase due transmit power control commands that are send
to the amount of time and resources already to the MD as part of the downlink control sig-
involved in the offloading data transmission, the nalling [16]. However, the above power control
avoidance of dropping a download connection for mechanisms are based on an important assump-
offloading should be considered with a higher pri- tion that there is a correlation between being
ority. Finally, the interference introduced to the close to the serving cell and being far away from
neighboring cells should be considered in the AC neighbor cells, which may not be true in HetNet.
and CA functions. Therefore, if the macrocells and small cells are
co-channel deployed, the varying uplink interfer-
POWER CONTROL AND RESOURCE ALLOCATION ence across cells may impact the effectiveness of
Through cooperation between different kinds of the FPC scheme. As a result, uplink power con-
radio access nodes, HetNet can perform power trol in HetNet has been studied in recent years.
control and resource allocation of multiple cells For example, the concept of outgoing interfer-
simultaneously to achieve maximum throughput ence based power control has been proposed,
and meet the power and QoS requirements for where a MD generating high interference to
MCC applications. In HetNet, many wireless other cells should transmit at low power, and
links potentially interfere with each other vice versa. Also, a MD can also adjust its trans-
because their allocated resources are not fully mit power based on the incoming interference,
orthogonal in order to better reuse the limited and transmit at high power level if it is subject to
radio resources. For example, interference sce- serious interference [17].
narios of femtocell can be considered as a spe-
cial case of that of HetNet. The interference Resource Allocation — Wireless network scheduling
between macrocell and femtocell, i.e., inter-tier is an important resource allocation function.
interference, arises from the fact that femtocells Since the MCC applications are usually delay
may utilize the spectrum already allocated to the and power sensitive, the scheduling function
macrocell. Meanwhile, all the femtocells can should be designed with the corresponding
share the same radio resources for improving objectives. Most research on scheduling has been
the resource usage, which may cause the inter- treating the downlink scenario with the following
ference between femtocells themselves, i.e., optimization objectives. Some comprehensive
intra-tier interference. Without proper interfer- surveys are provided in [18].
ence management methods, significant power is Maximum throughput and fairness tradeoff.
likely to be wasted in order to maintain an The channel variation due to the fast fading
acceptable user performance. For example, high effects of wireless channel is exploited as an
transmit power is usually radiated from a cell opportunity by the scheduler, which tends to
edge outdoor MD associated with a macrocell choose the MD with the best channel condition
eNodeB to provide reliable communications. If for transmission at every time slot. However, it
no proper interference management method is should also provide a fair share of resources to
applied, interference is possibly generated to all the MDs in the long term.
nearby indoor MDs connected to a femtocell Queue stability. The aim of the scheduler is
eNodeB in case that the whole or a part of fre- to ensure the queues’ stability without any
quency band is shared between the femtocell knowledge of arrival and channel statistics if
and macrocell. Thereupon, the femtocell MDs indeed stability can be achieved under any poli-
have to increase their transmission power to cy. Therefore, the scheduler takes into not only
maintain the communication with their indoor the channel state but also the queue state, and
the examples are MaxWeight and Exponential above power control, scheduling and interfer- With the rapid develop-
rule, etc. ence coordination methods in (heterogeneous)
Minimum delay or queue overflow probabili- wireless networks are inadequate. The schedul- ment of HetNet, the
ty. Since stability is a weak form of performance ing algorithms should consider the delay and
optimality, some research work focus on schedul- energy efficiency tradeoff for the support of
mobile users may enjoy
ing polices that minimize the overall average MCC applications. However, the existing studies the cloud services with
delay (per data unit) seen by the MDs; or mostly focus on the downlink scenario while the
scheduling policies which minimize the probabil- MCC application cares about the power con- good user experience
ity that either the sumqueue or the largest queue sumed by the MD instead of the eNodeB. There- regardless of spectrum
overflows a large buffer, such as the Log rule. fore, an uplink scheduling algorithm with delay
Energy efficiency and delay tradeoff. Energy and power guarantee has to be designed, where scarcity. For this pur-
efficiency is an important concern in the design the contiguous allocation constraint has to be
of modern wireless systems. However, it is often considered. Furthermore, considering those
pose, we propose a
achieved at the cost of increasing delay due to MDs in the wireless networks without MCC framework of HetNet for
two reasons. First, since the required transmis- applications, the scheduling algorithm should
sion power is a convex function of the communi- consider the mix traffic scenario. Finally, the MCC, which is based on
cation rate, this implies that transmitting data at scheduling algorithm should be optimized jointly the power and QoS
low rates over a longer duration is more energy with the power control and interference coordi-
efficient as compared to high rate transmissions. nation functions in HetNet to achieve the maxi- negotiation between the
Second, since the wireless channel is time-vary- mum spectrum efficiency while guaranteeing the
ing and as good channel conditions require less power and QoS requirements for the offloading
offloading decision mod-
transmission power, scheduling MDs only at connections. ule and the wireless
good channel conditions leads to reduced energy
cost. Therefore, the problem of minimizing ener- CONCLUSION HetNet.
gy subject to delay constraint or minimizing
delay subject to energy constraints are treated HetNet consisting of macrocells and small cells is
for single user [19] and multiuser wireless sys- expected to provide the wireless connection any-
tems [20]. where and anytime. With the rapid development
A mixture of optimization objectives for dif- of HetNet, the mobile users may enjoy the cloud
ferent traffic. Since there is usually a mixture of services with good user experience regardless of
elastic and real-time traffic in realistic wireless spectrum scarcity. For this purpose, we propose a
networks, several studies that addresses the framework of HetNet for MCC, which is based
problem of simultaneously supporting these traf- on the power and QoS negotiation between the
fic have been studied, ensuring that the real-time offloading decision module and the wireless Het-
traffic receiving their desired QoS while the elas- Net. However, several challenges have to be
tic traffic achieves the maximum possible addressed before the users can truly enjoy the
throughput. life from MCC applications, which are discussed
Compared with the large amount of research in the article. Future research on HetNet for
on downlink scheduling, considerably less work MCC will be conducted based on a tight coupling
has been dedicated to the uplink. LTE uses the of the unique characteristics of the MCC applica-
SC-FDMA radio access technology for its uplink tions and the wireless heterogeneous networks.
transmission. As a result, resources assigned to Specifically, the computation offloading decision
the same MD must be contiguous in the fre- should be made considering the power and QoS
quency domain. This contiguous allocation con- guarantee from the HetNet along with the cost
straint limits the scheduling flexibility and makes and refunded penalty rates. The wireless HetNet,
the above optimization problems more complex on the other hand, should apply the related radio
for the uplink scenario. As the optimal solutions resource management functions, e.g., admission
would mostly be NP-hard, the proposed schedul- control and resource allocation, to provide the
ing algorithms are often based on heuristics power and QoS guarantee of the MCC applica-
yielding reasonable system performance under tions whenever feasible.
practical circumstances [21].
The resource allocation function should also ACKNOWLEDGMENT
consider interference coordination by allocating This work was supported by the National Natu-
different resources between neighboring ral Science Foundation of China (No. 61272168),
eNodeBs in the time or frequency domains in Program for New Century Excellent Talents in
order to mitigate co-channel interference. The University (NCET-11-0600), the Fundamental
main challenge lies in the fact that the location Research Funds for the Central Universities
and coverage areas of small cells such as femto- (No. 2012JBM003), the State Key Laboratory of
cells are uncertain, and the traffic load in small Rail Traffic Control and Safety (Contract No.
cells are less aggregated due to the much fewer RCS2011ZT005), Beijing Jiaotong University,
served MDs per cell compared to the macro and the Key Grant Project of Chinese Ministry
cells. The coordination requires communication of Education (No. 313006).
between different network nodes in order to
(re)configure radio resources. Based on the REFERENCES
needs of the inter-sites communication interval, [1] H. T. Dinh et al., “A Survey of Mobile Cloud Computing:
most interference coordination schemes may be Architecture, Applications, and Approaches,” Wireless
categorized into two classes, i.e., semi-static and Commun. and Mobile Computing, Oct 2011.
[2] K. Zheng et al., “Interference Coordination for OFDM-
dynamic schemes. based Multihop Cellular Networks Towards LTE-
When considering the support of MCC appli- Advanced,” IEEE Wireless Commun., vol. 18, no. 1, Feb.
cations with power and QoS constraints, the 2011, pp. 54–63.
[3] K. Zheng et al., “Energy-Efficient Wireless In-Home: the [20] M. J. Neely, “Optimal Energy and Delay Tradeoffs for
Need for Interference-Controlled Femtocells,” IEEE Multi-User Wireless Downlinks,” IEEE Trans. Info. Theo-
Wireless Commun., vol. 18, no. 6, Dec. 2011, pp. ry, vol. 53, no. 9, Sept. 2007, pp. 1–17.
36–44. [21] S.-B. Lee et al., “Proportional Fair Frequency-Domain
[4] Y. Ye et al., “A Framework for QoS and Power Manage- Packet Scheduling for 3GPP LTE Uplink,” IEEE Infocom
ment in A Service Cloud Environment with Mobile 2009, pp. 2611–15.
Devices,” IEEE Int’l. Symp. Service Oriented System
Engineering (SOSE), 2010, June 2010, pp. 236–43. BIOGRAPHIES
[5] M. Abundo, V. Cardellini, and F. Lo Presti, “An MDP-
LEI LEI received a B.S. degree in 2001 and a Ph.D. degree in
based Admission Control for a QoS-Aware Service-Ori-
2006, respectively, from Beijing University of Posts &
ented System,” 2011 IEEE 19th Int’l. Wksp. Quality of
Telecommunications, China, both in telecommunications
Service (IWQoS), June 2011, pp. 1–3.
engineering. From July 2006 to March 2008, she was a
[6] E. Cuervo et al., “MAUI: Making Smartphones Last
postdoctoral fellow at Computer Science Department,
Longer with Code Offload,” Proc. 2010 Int’l. Conf.
Tsinghua University, Beijing, China. She worked for the
Mobile Syst., App., Services, 2010, pp. 49–62.
Wireless Communications Department, China Mobile
[7] B. G. Chun et al., “CloneCloud: Elastic Execution
Research Institute from April 2008 to August 2011. She
Between Mobile Device and Cloud,” Proc. 6th Conf.
has been an Associate Professor with the School of Elec-
Computer Systems (EuroSys), Apr. 2011, pp. 301–14.
tronic and Information Engineering, Beijing Jiaotong Uni-
[8] S. Gitzenis and N. Barnbos, “Joint Task Migration and
versity since September, 2011. Her current research
Power Management in Wireless Computing,” IEEE
interests include performance evaluation, quality-of-service
Trans. Mobile Computing, vol. 8, no. 9, Sept. 2009, pp.
and radio resource management in wireless communica-
1189–204.
tion networks.
[9] Y. Wen, W. Zhang, and H. Luo, “Energy-Optimal Mobile
Application Execution: Taming Resource-Poor Mobile
K AN Z HENG [SM’09] (kzheng@ieee.org) received the B.S.,
Devices with Cloud Clones,” Proc. IEEE Infocom, 2012,
M.S. and Ph.D. degree from Beijing University of Posts &
pp. 2716–20.
Telecommunications (BUPT), China, in 1996, 2000 and
[10] K. Kumar and Y.-H. Lu, “Cloud Computing for Mobile
2005, respectively, where he is currently associate profes-
Users: Can Offloading Computation Save Energy?,”
sor. He worked as a researcher in the companies including
Computer, vol. 43, no. 4, Apr. 2010, pp. 51–56.
Siemens, Orange Labs R & D (Beijing), China. His current
[11] D. Huang, P. Wang, and D. Niyato, “A Dynamic
research interests lie in the field of wireless communica-
Offloading Algorithm for Mobile Computing,” IEEE
tions, with emphasis on heterogeneous networks and
Trans. Wireless Commun., vol. 11, no. 6, June 2012,
M2M networks.
pp. 1991–95.
[12] E. Z. Tragos, G. Tsiropoulos, G. T. Karetsos, and S. A.
JIADI CHEN received her B.S. degree from Chongqing Univer-
Kyriazakos, “Admission Control for QoS Support in Het-
sity & Posts and Telecommunications (CUPT), China, in
erogeneous 4G Wireless Networks,” IEEE Network,
2010. She is currently a candidate for Ph.D. in the Key Lab
May/June 2008, pp. 30–37.
of Universal Wireless Communications, Ministry of Educa-
[13] L. B. Le et al., “Joint Load Balancing and Admission
tion, BUPT. Her research interests include performance
Control in OFDMA-Based Femtocell Networks,” IEEE ICC
analysis of wireless networks, resource allocation and
2012, 2012, pp. 5135–39.
scheduling algorithm.
[14] S. Corroy, L. Falconetti, and R. Mathar, “Cell Associa-
tion in Small Heterogeneous Networks: Downlink Sum
HANLI MENG received her B.S. degree from Beijing University
Rate and Min Rate Maximization,” IEEE WCNC 2012,
of Posts & Telecommunications (BUPT), China, in 2013. She
2012, pp. 898–902.
is currently a candidate for M.S in the Key Lab of Universal
[15] Y. Zhang et al., “Performance Analysis of user Associa-
Wireless Communications, Ministry of Education, BUPT. Her
tion Policies in Small Cell Networks using Stochastic
research interests include resource allocation and schedul-
Petri Nets,” IEEE ICC Wksp. 2013.
ing algorithm in wireless networks.
[16] A. Simonsson, and A. Furuskar, “Uplink Power Control
in LTE C Overview and Performance,” IEEE 68th Vehic.
Z HANGDUI Z HONG received the B.Eng. and M.Sc. degrees
Tech. Conf., VTC 2008-Fall, Sept. 2008, pp. 1–5.
from Northern Jiaotong University (currently Beijing Jiao-
[17] Y. Wang, and S. Venkatraman, “Uplink Power Control
tong University), Beijing, China, in 1983 and 1988, respec-
for Heterogeneous Networks in LTE,” 2012 IEEE Globe-
tively. He has been a Professor with the School of Electronic
com Workshops (GC Wksps.), Dec. 2012, pp. 592–97.
and Information Engineering, Beijing Jiaotong University,
[18] M. Andrews, “A Survey of Scheduling Theory in Wire-
since 2000. He has authored seven books and over 150
less Data Networks,” Wireless Communications (The
technical papers in the field of wireless communication. His
IMA Volumes in Mathematics and Its Applications), vol.
current research interests include wireless communication
143, 2007, pp. 1–17.
theory for railway systems, wireless ad hoc networks, chan-
[19] J. Lee and N. Jindal, “Energy-Efficient Scheduling of
nel modeling, radio resource management, intelligent
Delay Constrained Traffic over Fading Channels,” IEEE
transportation systems, and GSM-R systems.
Trans. Wireless Commun., vol. 8, no. 4, pp. 1866–75,
Apr. 2009.