Security-aware 5G RAN slice mapping with tiered&#x00A0;isolation in physical-layer secured metro-aggregation elastic optical networks using&#x00A0;heuristic-assisted DRL

Yunwu Wang; Yunwu Wang; Min Zhu; Min Zhu; Jiahua Gu; Jiahua Gu; Xiang Liu; Xiang Liu; Weidong Tong; Weidong Tong; Bingchang Hua; Mingzheng Lei; Yuancheng Cai; Jiao Zhang

doi:10.1364/JOCN.499551

Journal of Optical Communications and Networking
Vol. 15,
Issue 12,
pp. 969-984
(2023)
•https://doi.org/10.1364/JOCN.499551

Security-aware 5G RAN slice mapping with tiered isolation in physical-layer secured metro-aggregation elastic optical networks using heuristic-assisted DRL

Yunwu Wang, Min Zhu, Jiahua Gu, Xiang Liu, Weidong Tong, Bingchang Hua, Mingzheng Lei, Yuancheng Cai, and Jiao Zhang

Not Accessible

Your library or personal account may give you access

Get PDF
Email
Share
Get Citation
Copy Citation Text
Yunwu Wang, Min Zhu, Jiahua Gu, Xiang Liu, Weidong Tong, Bingchang Hua, Mingzheng Lei, Yuancheng Cai, and Jiao Zhang, "Security-aware 5G RAN slice mapping with tiered isolation in physical-layer secured metro-aggregation elastic optical networks using heuristic-assisted DRL," J. Opt. Commun. Netw. 15, 969-984 (2023)

Export Citation
- BibTex
- Endnote (RIS)
- HTML
- Plain Text
Citation alert
Save article

Abstract

The optical transport network (OTN) encryption technology is attractive to solve the physical-layer security in services for the light-path provision process. This paper mainly explores the security-aware 5G radio access network (RAN) slice mapping problem with the tiered isolation (TI) policy, which decides the solution for aggregating service into the physical-layer secured metro-aggregation elastic optical networks (MA-EONs). We first introduce the physical-layer secured OTNs and illustrate their differences from the traditional optical networks. Then, we formulate the 5G RAN slice mapping problem in physical-layer secured MA-EONs as an exact integer linear programming (ILP) model to minimize the average cost (AC), which consists of the number of utilized processing pools (PPs)/general-purpose processors (GPPs)/virtual machines (VMs), and maximum frequency slot index (MFSI) on the light-paths, meanwhile satisfying the given slice’s latency, isolation, and security requirements. After that, to overcome the non-scalability problem of the ILP model, a heuristic-assisted deep reinforcement learning (HA-DRL) algorithm is proposed to obtain a near-optimal solution for large-scale network scenarios, where the classical shortest path algorithm is employed in the DRL to shrink the size of the exploration space and accelerate the convergence process. Finally, we evaluate the proposed ILP model and HA-DRL algorithm through extensive simulations. Simulation results indicate that our proposed HA-DRL method can find approximate solutions to the ILP model in the small-scale network scenario. Furthermore, the HA-DRL method can also achieve higher resource efficiency compared with benchmark heuristic first-fit algorithms in the large-scale network scenario. In comparison to the first-fit algorithm benchmark, the proposed HA-DRL can achieve up to 9.4% AC reduction in large-scale network scenarios.

Full Article | PDF Article

More Like This

Resource-efficient and QoS guaranteed 5G RAN slice migration in elastic metro aggregation networks using heuristic-assisted deep reinforcement learning

Jiahua Gu, Min Zhu, Yunwu Wang, Xiaofeng Cai, Yuancheng Cai, Jiao Zhang, Mingzheng Lei, Bingchang Hua, Pingping Gu, and Guo Zhao
J. Opt. Commun. Netw. 15(11) 854-870 (2023)

Mapping sub-wavelength service with heterogenous security demand over physical-layer secured optical transport networks

Tianhe Liu, Wei Wang, Xiangkun Man, Huifang Xiong, Yongli Zhao, Guangquan Wang, Yajie Li, Yanxia Tan, Shan Dong, and Jie Zhang
J. Opt. Commun. Netw. 15(5) 268-281 (2023)

DRL-assisted joint allocation of antenna, radio, and front-haul resources in TWDM-PON-based NG-RANs with mMIMO-enabled beamforming

Min Zhu, Yunwu Wang, Jiahua Gu, Xiaofeng Cai, Xiang Liu, Weidong Tong, Yuancheng Cai, and Jiao Zhang
J. Opt. Commun. Netw. 15(5) 241-254 (2023)

Cited By

You do not have subscription access to this journal. Cited by links are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Figures (6)

You do not have subscription access to this journal. Figure files are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Tables (10)

You do not have subscription access to this journal. Article tables are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Equations (46)

You do not have subscription access to this journal. Equations are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Abbreviation	Full Form	Abbreviation	Full Form
5G	Fifth generation	ILP	Integer linear programming
AAU	Active antenna unit	LSL	Link security level
AC	Average cost	MA-EON	Metro-aggregation elastic optical network
BBU	Baseband unit	MCS	Modulation and coding scheme
BV-OXC	Bandwidth-variable optical cross-connect	MFSI	Maximum frequency slot index
C-RAN	Centralized radio access network	MIMO	Multiple input multiple output
CT	Constraint	mMIMO	massive multiple input multiple output
CU	Centralized unit	NFV	Network function virtualization
DC	Data center	NG-RAN	Next generation radio access network
DDQN	Double deep Q network	OEO	Optical–electric–optical
DDPG	Deep deterministic policy gradient	OTN	Optical transport network
DNN	Deep neural network	OS	Operating system
DRL	Deep reinforcement learning	PP	Processing pool
DU	Distributed unit	RB	Resource block
EON	Elastic optical network	SDN	Software defined network
E-switch	Ethernet switch	SF	Service function
FF	First-fit	SFC	Service function chain
FFA	First-fit algorithm	SRA	Shortest path random algorithm
FS	Frequency slot	TI	Tiered isolation
GOPS	Giga operations per second	UPF	User plane function
GPP	General-purpose processor	VM	Virtual machine
GNN	Graph neural network	VNF	Virtual network function
HA-DRL	Heuristic-assisted deep reinforcement learning	WDM	Wavelength division multiplexing

Reference	Isolation	Security	ILP	Heuristic	DRL
[18]	$\sqrt{}$		$\sqrt{}$	$\sqrt{}$
[19]	$\sqrt{}$			$\sqrt{}$
[20]	$\sqrt{}$		$\sqrt{}$	$\sqrt{}$
[21]			$\sqrt{}$	$\sqrt{}$	$\sqrt{}$
[22]			$\sqrt{}$		$\sqrt{}$
[23]			$\sqrt{}$	$\sqrt{}$	$\sqrt{}$
[24]	$\sqrt{}$		$\sqrt{}$		$\sqrt{}$
This	$\sqrt{}$	$\sqrt{}$	$\sqrt{}$	$\sqrt{}$	$\sqrt{}$

Notation	Description
$B, L$	Set of AAUs and optical links
$R, G$	Set of PPs and GPPs in each PP
$V, F$	Set of VMs in each GPP and FSs in each link
$S$	Set of SFs (e.g., $s = 1$ for AAU, $s = 2$ for DU, $s = 3$ for CU, $s = 4$ for metro DC)
$E_{r}$	Equals 1 if metro DC is deployed at PP $r$
$M_{i, j}$	Equals 1 if PP $i$ is directly connected to PP $j$
$F_{b, r}$	Equals 1 if AAU $b$ is directly connected to PP $r$
$C_{r, g, v}$	Computing capacity of each VM $v$ in GPP $g$ at PP $r$
$T_{b}$	End-to-end latency requirement of AAU $b$
$τ_{b}$	Front-haul latency requirement of AAU $b$
${T C}_{b}$	Transmission latency from AAU $b$ to its directly connected PP
$T_{o e o}$	Latency of OEO operation
$T_{i, j}$	Propagation latency of link $l (i, j)$ , $i, j \in R$
$Q_{i, j}$	LSL of link $l (i, j)$ , $i, j \in R$
${C K}_{b, s}$	Computing demand of SF $s$ of AAU $b$
${T I}_{b, s}$	TI level demand of SF $s$ of AAU $b$
${R B}_{b, s}$	FSs demand of SF $s$ of AAU $b$
${L S}_{b, s}$	LSL demand of AAU $b$ with the previous SF $s$ being processed
Num	Large positive integer

Variable	Description
$D_{r}$	Equals 1 if PP $r$ is activated
$Γ_{r, g}$	Equals 1 if GPP $g$ at PP $r$ is activated
$Δ_{r, g, v}$	Equals 1 if VM $v$ in GPP $g$ at PP $r$ is activated
$H_{b, r}$	Equals 1 if AAU $b$ is processed in PP $r$
$Θ_{b, r, g}$	Equals 1 if AAU $b$ is processed in GPP $g$ at PP $r$
$Λ_{b, r, g, v}$	Equals 1 if AAU $b$ is processed in VM $v$ of GPP $g$ at PP $r$
$U_{b, s, r}$	Equals 1 if SF $s$ of AAU $b$ is processed in PP $r$
$Ξ_{b, s, r, g}$	Equals 1 if SF $s$ of AAU $b$ is processed in GPP $g$ at PP $r$
$O_{b, s, r, g, v}$	Equals 1 if SF $s$ of AAU $b$ is processed in VM $v$ of GPP $g$ at PP $r$
$Ω_{i, j, f}$	Equals 1 if FS $f$ of link $l (i, j)$ is used
$X_{b, s, i, j}$	Equals 1 if AAU $b$ is carried on the link $l (i, j)$ with the previous SF $s$ being processed
$Ψ_{b, s, i, j, f}$	Equals 1 if AAU $b$ is carried on the FS $f$ on the link $l (i, j)$ with the previous SF $s$ being processed
$P_{i, j}$	Equals 1 if link $l (i, j)$ is used
MFSI	Number of the maximum FS index in physical-layer secured MA-EONs

	Parameter	Small-Scale Network	Large-Scale Network
Wireless part	Modulation bits	6	6
	Coding rate	772/1024	772/1024
	Number of antennas	4	4
	Number of MIMO layers	2	2
	Number of physical RBs	500	500
MA part	Number of GPPs in each PP	2	10
	Number of VMs in each GPP	2	10
	VM capacity	4000 GOPS	4000 GOPS
	TI levels	TI3	TI0, TI1, TI2, TI3
Light-path part	Each optical link ranges	[5,30] km	[10,30] km
	Bandwidth capacity for each FS	12.5 Gbps	12.5 Gbps
	LSL requirements	[1,2]	[1,2,3,4]
	Number of MFSI	15	358

	Number of AAUs
Algorithm	40	50	60	70	80	90	100
FFA	0.33	0.59	0.78	0.97	1.15	1.37	1.55
SRA	0.12	0.16	0.19	0.22	0.25	0.26	0.30
HA-DRL	4.03	8.60	13.3	21.2	30.5	42.7	55.2

Abstract

Cited By

Figures (6)

Tables (10)

Equations (46)

Journal of Optical Communications and Networking