Joint Communication and Computation in Hybrid CloudMobile Edge Computing Networks Robert-Jeron Reifert Hayssam Dahroujy Basem Shihadaz Aydin Sezgin

2025-05-05 0 0 385.49KB 7 页 10玖币

侵权投诉

Joint Communication and Computation in Hybrid

Cloud/Mobile Edge Computing Networks

Robert-Jeron Reifert∗, Hayssam Dahrouj†, Basem Shihada‡, Aydin Sezgin∗,

Tareq Y. Al-Naffouri‡and Mohamed-Slim Alouini‡

∗Institute of Digital Communication Systems, Ruhr-Universit¨

at Bochum, Germany

†Department of Electrical Engineering, University of Sharjah, Sharjah, United Arab Emirates

†Communication Theory Lab, King Abdullah University of Science and Technology, Thuwal, Saudi Arabia

Abstract—Facing a vast amount of connections, huge per-

formance demands, and the need for reliable connectivity, the

sixth generation of communication networks (6G) is envisioned

to implement disruptive technologies that jointly spur connec-

tivity, performance, and reliability. In this context, this paper

proposes, and evaluates the beneﬁt of, a hybrid central cloud

(CC) computing and mobile edge computing (MEC) platform,

especially introduced to balance the network resources required

for joint computation and communication. Consider a hybrid

cloud and MEC system, where several power-hungry multi-

antenna unmanned aerial vehicles (UAVs) are deployed at the

cell-edge to boost the CC connectivity and relieve part of its

computation burden. While the multi-antenna base stations are

connected to the cloud via capacity-limited fronthaul links,

the UAVs serve the cell-edge users with limited power and

computational capabilities. The paper then considers the problem

of maximizing the weighted network sum-rate subject to per-

user delay, computational capacity, and power constraints, so as

to determine the beamforming vectors and computation alloca-

tions. Such intricate non-convex optimization problem is tackled

using an iterative algorithm that relies on `0-norm relaxation,

successive convex approximation, and fractional programming,

and has the compelling ability to be implemented in a distributed

fashion across the multiple UAVs and the CC. The paper results

illustrate the numerical prospects of the proposed algorithm for

enabling joint communication and computation, and highlight

the appreciable improvements of data processing delays and

throughputs as compared to conventional system strategies.

I. INTRODUCTION

Today’s Internet of Things (IoT) applications involve many

relevant consumer and industry use cases, e.g., smart cities,

modular plant, etc. For the next generation of wireless com-

munication systems, massive IoT is forecasted to make up

of 51% cellular IoT connections by 2027, while other use-

cases extend to augmented reality, vehicular to anything, etc.

[1]. With massively increased number of connected devices

and increased service requirements, massive IoT raises further

challenges towards the realization of the sixth generation of

communication networks (6G). Due to their ﬁnite energy and

computation resources, IoT devices are often dependent on

ofﬂoading their tasks, especially for computation-intensive

applications [2]. A suitable technique for satisfying such

massive data demand is the cloud-based network architecture,

1This work was was supported in part by the German Federal Ministry of

Education and Research (BMBF) in the course of the 6GEM Research Hub

under Grant 16KISK037.

which enables centralized management of communication and

computation resources. However, a drawback of cloud-based

networks is the long propagation delay [3] and the need for

costly, limited-capacity fronthaul links to connect the cloud to

the base stations (BSs). Moving computation and management

capabilities towards the network edge enables both latency

reduction and service quality enhancement. Such cost-efﬁcient

and energy-saving paradigm, referred to as mobile edge com-

puting (MEC), is subject, however, to strict constraints on

power and computational resources. To this end, this paper

considers one particular hybrid network architecture, where

the central cloud connects to central BSs so as to serve

the central network users. The cell-edge users, on the other

hand, are served by resource-limited MEC devices, especially

deployed to boost the system connectivity. The paper then

adopts such a hybrid cloud/MEC architecture to empower joint

communication and computation by means of maximizing a

network-wide sum-rate, the performance of which is a function

of the allocated computation and communication resources.

The topic considered in this paper is related to the general

context of resource management in cloud-radio access net-

works [4], [5], and MEC-based works focusing on computa-

tion [2], [6], [7] and communication [3], [8], [9]. Unmanned

aerial vehicles (UAV)-assisted MEC has been recognized

as a promising 6G network technique for allowing ﬂexible

deployment, on-demand service, and enhanced connectivity

[3], [10], [11]. The utilization of UAVs, with typically strict

power constraints, calls for sophisticated joint management

of communication and computation resources in order to

optimize the system performance. Due to weak received power

and strong adjacent network interference, especially in the

prospective 6G ultra-dense deployment, cell-edge users are

often prone to inferior service quality, which makes the

development of communication and computation resource

management techniques vital. Related works in this ﬁeld

include UAV-aided communication [8], [9] and computation

[6], [12] networks. These works, however, do not capture

the joint consideration of the communication and computation

aspects, i.e., constraint-wise and variable-wise, e.g., see [10].

As 6G communication networks are envisioned to include

multiple access technologies in a hybrid manner, the need to

consider the interplay of a central network and edge devices

arXiv:2210.02090v1 [cs.IT] 5 Oct 2022

[10], [11], [13], this work extends the joint communication

and computation paradigm toward hybrid cloud and MEC

networks. Further, in contrast to previous works on MEC

networks which adopt orthogonal access schemes, e.g., [14],

[15], this paper adopts a spatial multiplexing approach and

separates users using a beamforming strategy. Reference [4]

utilizes a similar approach for resource management under the

multi-cloud paradigm; however, the essential computation and

delay considerations are ignored in [4] which rather focuses

on mitigating the intra- and inter-cloud interference in a multi-

cloud setup in the absence of any MEC capabilities.

Unlike the aforementioned references, this paper proposes

a downlink hybrid cloud/MEC network, where several multi-

antenna BSs and UAVs serve single-antenna network users.

The BSs are connected to the cloud via capacity limited

fronthaul links, while the UAVs perform computation and

communication functions on their own. We address a sum-

rate maximization problem by jointly managing beamforming

vectors, allocated rates, and computation capacity, subject to

per-BS and per-UAV power, per-BS fronthaul capacity, per-

computing platform maximum computation capacity, and per-

user delay constraints. Such mixed discrete-continuous non-

convex optimization problem is tackled using `0-norm relax-

ation, successive convex approximation (SCA), and fractional

programming (FP) resulting in a fully centralized protocol

(FCP) and in an efﬁcient partially decentralized protocol

(PDP). Insightful simulations verify the gains of the proposed

network architecture in terms of sum-rate and delay. The

proposed decentralized algorithm is particularly shown to

overcome the centralized version in terms of computational

complexity, runtime, and scalability, as well as the fully

distributed protocol (FDP) in terms of sum-rate.

II. SYSTEM MODEL AND PROBLEM FORMULATION

In this work, we consider the downlink of a hybrid

cloud/MEC-based network architecture. Under such frame-

work, cloud processors (CPs) coordinate the users operations

within the core-network. The UAVs, on the other hand, with

on-chip computation capabilities, act as mobile edge com-

puters (ECs) to serve the cell-edge users. The core network

consists of a single cloud, i.e., the central cloud (CC), con-

nected via fronthaul links to Bmulti-antenna BSs, with Lc

antennas each, while the UAVs at the edge are equipped with

Leantennas each. The split of network functions follows a

data-sharing approach, where the CP at the CC performs most

network functions, e.g., encoding and precoder design, leaving

the modulation, precoding, and radio tasks to the BSs [5].

Fig. 1 shows an example of the considered system, which

illustrates a network of 2BSs serving 4central users, and 2

UAVs each serving one user. Let Ebe the number of deployed

ECs, and let E={1,· · · , E}be the set of ECs. Since each

EC is implemented on a UAV, the edge network consists of E

UAVs. Note that throughout this work, the terms ECs and

UAVs are interchangeably used without loss of generality.

The set of BSs is given by B={1,· · · , B}. The set of

single-antenna users is denoted by K={1,· · · , K}, where

Communication link

Cloud-UAV interference

Intra-cloud interference

Coordination link

User

CP UAV

Base station

Fig. 1: Network of 2UAVs, 2BSs, and 6users.

Kis the total number of users. In the context of CC and EC

coexistence, this paper assumes disjoint user-clusters, which

are covered by the user sets Kc⊆ K and Ke⊆ K, with

Kc∩ Ke=∅,∀e∈ E and Ke∩ K0

e=∅,∀e6=e0. In other

terms, the set of users served by the CC is denoted by Kc,

while the set of users served by each EC eis denoted by Ke,

∀e∈ E. Similarly, the CC serves Kcusers, while EC eserves

Keusers. The determination of the sets Keand Kcfalls outside

the scope of the current paper, as it is often determined on a

different time-scale than the beamforming problem considered

in this paper.

We further denote the channel vector from BS bto user

kby hb,k ∈CLc, and the channel vector from UAV eto

user kby ˜

he,k ∈CLe, where e∈ E. The aggregate channel

vector from all BSs and UAVs towards user kis given by

hk= [hT

1,k,· · · ,hT

B,k,˜

h1,k,· · · ,˜

hE,k]T. For mathematical

tractability, the paper assumes full knowledge of channel state

information at the transmitters (CSIT).

The paper then aims at jointly managing the communica-

tion and computation network resources, which are captured

through the following list of variables:

•The beamforming vectors wb,k ∈CLcand ˜

we,k ∈CLe,

which denote the beamforming vector of user k’s signal

at BS band at UAV e∈ E, respectively.

•The computation vector f∈NK, which denotes the

allocated computation cycles to process the users data,

where each entry fkis given in cycles/s, ∀k∈ K.

•The rate allocation vector r∈RK, which denotes the

achievable rates of all users, with r= [r1,· · · , rK].

Note that wk= [wT

1,k,· · · ,wT

B,k,˜

1,k,· · · ,˜

E,k]Tis the

aggregate beamforming vector for user k. As per the user

association constraints, wkis group-sparse by design since k

may only be served by either CC or one EC. Additionally, in

the case when the CC serves the user k, not all BSs participate

in serving k, i.e., wb,k =0Lcfor some BSs b. We next

describe the expressions of the metrics relevant to the paper

context, mainly, the data-rates, the power consumption at the

BSs and at the UAVs, the users transmission delays, and the

computation capacity.

a) Achievable Rate: Each user receives its own intended

signal h†

kwksk, and treats all other users’ signals as interfer-

ence. This is captured in the signal to interference plus noise

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

JointCommunicationandComputationinHybridCloud/MobileEdgeComputingNetworksRobert-JeronReifert,HayssamDahroujy,BasemShihadaz,AydinSezgin,TareqY.Al-NaffourizandMohamed-SlimAlouinizInstituteofDigitalCommunicationSystems,Ruhr-Universit¨atBochum,GermanyyDepartmentofElectricalEngineering,UniversityofSha...

展开>> 收起<<

Joint Communication and Computation in Hybrid CloudMobile Edge Computing Networks Robert-Jeron Reifert Hayssam Dahroujy Basem Shihadaz Aydin Sezgin.pdf

共7页,预览2页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

Joint Communication and Computation in Hybrid CloudMobile Edge Computing Networks Robert-Jeron Reifert Hayssam Dahroujy Basem Shihadaz Aydin Sezgin

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: