Availability-Guaranteed Slice Provisioning in Wireless-Optical

Broadband Access Networks Supporting Mobile Edge Computing

Gangxiang Shen

School of Electronic and Information Engineering Soochow University, China

Acknowledgment: Ke Chen, Shuiping Jie, Boping Jiang, Sanjay K. Bose

Outline

Background

New Definition for Availability

Our Research Problem

Heuristic Algorithms

Test Conditions

Simulations and Performance Analyses

Conclusions

2

Introduction to 5G

3

5G2G 3G 4G

Faster and faster

Ultra-Low latency

Ultra-large-scale access

uRLLC

Ultra-Reliable Low latency Communications

mMTC

massive Machine Type communications

eMBB

enhanced Mobile Broadband

Application

8K

Autopilot

8K

Smart home

5G Key Requirements

4Source: Delta Partners Analysis

X-Haul for 5G

5

Front-haul Mid-haul Back-haul

Source: OFC 2018, Tu2K.1

–- <MEC Deployment in 4G

and Evolution towards 5G>

ETSI White Paper

Augmented Reality Service

Scenario

–- <Mobile Edge Computing A

Key technology towards 5G>

ETSI White Paper

Mobile Edge Computing (MEC)

6

Base stationMEC server

GW-UP

GW-CP

NetworkSlice

1

NetworkSlice

2

NetworkSlice

n

CN FireWall

Internet

RAN

MEC: Mobile Edge Computing RAN: Radio Access Network

Wireless-Optical Broadband AccessNetwork (WOBAN) with MEC

7

OLT

ONU-BS

ONU-BS

ONU-BS

OLT: Optical Line Terminal

RN: Remote Node

ONU-BS: Optical Network Unit-

Base Station

UE: User Equipment

ONU

VM

BS

Fiber Link

Microwave Link

ONU-BS

RN

RN

OLT

UE

BS

BS

ONU-BS

ONU-BS

ONU-BS

ONU-BS

ONU-BS

ONU-BS

P. Chowdhury, B. Mukherjee, et al., IEEE Network 23(3), 41-48 (2009).

Composition: Passive Optical Network (PON) + Wireless Mesh

Network (WMN)

Advantages: High bandwidth, stability, low cost, and flexibility

MEC

MEC

MEC

Network Virtualization and Slicing

8

A slice is considered as an independent network, consisting

of multiple virtual nodes and virtual links

A slice accommodates an independent application

Advantages: efficiency, flexibility

VN

VN

VNVN

PN

PN

VL

VL

FL

FLML

VN: Virtual Node

VL: Virtual Link

PN: Physical Node

FL: Fiber Link

ML: Microwave Link

Slice1

Slice2

WOBAN

ML

VN VNVL

Slice3eMBB

URLLC

mMTC

K. Samdanis, et al., IEEE Communications Magazine 54(7), 32-39 (2016).

Virtual node: Provide

computing/storage capacities

for supporting MEC

Virtual link: Provide network

bandwidth resources for

communication

Virtual Node and Link Mapping

9

VN

VN

PN

PN

VL

FL

FLML

VN: Virtual Node

VL: Virtual Link

PN: Physical Node

FL: Fiber Link

ML: Microwave Link

Slice

WOBAN

ML

K1

K2

v1

i

n

Each virtual node is embedded in a physical node

Each virtual link is mapped to one of physical paths

Objective: Efficiently share physical resources

5G Key Requirements

10Source: Delta Partners Analysis

Existing Studies

11

No studies considering guaranteed availability for each

provisioned slice in the context of a WOBAN

supporting MEC

Concept of network

slicing

Ordonez-Lucena J et al., IEEE Communications

Magazine.

Samdanis K et al., IEEE Communications Magazine.

Application of network

slicing

Mayoral A et al., 2016 ECOC . IEEE.

Lee Y L et al., IEEE Transactions on Wireless

Communications.

Efficiency of slice

provisioning

Trivisonno R et al., 2015 GLOBECOM. IEEE.

Zhang H et al., IEEE Communications Magazine.

P. Rost et al., Communications Magazine.

Availability of WOBAN Kiese M et al., 2009 IEEE International Conference

on Communications.

Shao X et al., 2010 OFC/NFOEC.

Availability-Guaranteed Slice Provisioning in Wireless-Optical Broadband Access Networks

Supporting Mobile Edge Computing

Outline

Background

New Definition for Availability

Our Research Problem

Heuristic Algorithms

Test Conditions

Simulations and Performance Analyses

Conclusions

12

Conventional Definition of Availability

𝐴 =𝑀𝑇𝑇𝐹

𝑀𝑇𝑇𝐹+𝑀𝑇𝑇𝑅

MTTF: Mean Time to Failure

MTTR: Mean Time to Repair

13

𝐴𝑖 =𝑀𝑇𝑇𝐹𝑖

𝑀𝑇𝑇𝐹𝑖+𝑀𝑇𝑇𝑅𝑖

𝐴 = 𝑖=1𝟒 𝐴𝑖

Component 2 Component 3 Component 4Component 1

MTTF MTTR

failurefailure

Definition of availability Availability of a serial system

Partially Functioning Slice

14

A single network failure would not cause all the virtual links to

fail; there would be a partial set of virtual links still functioning

VNVN

PN

PN

VL

FL

FLML

VN: Virtual Node

VL: Virtual Link

PN: Physical Node

FL: Fiber Link

ML: Microwave Link

Slice1

WOBAN

ML

failure

Still functioning!

New Definition for Availability

𝐴𝑠 =𝐶𝑛𝑜𝑟𝑚𝑎𝑙𝑠 + 𝐶𝑝𝑎𝑟𝑡𝑖𝑎𝑙

𝑠

𝐶𝑡𝑜𝑡𝑎𝑙𝑠 ∀𝑠 ∈ 𝑺

New ([0, 1.0])

15

MTTF MTTR

total

𝐶𝑡𝑜𝑡𝑎𝑙𝑠

𝐶𝑛𝑜𝑟𝑚𝑎𝑙𝑠 =1 0 ≤ 𝐶𝑝𝑎𝑟𝑡𝑖𝑎𝑙

𝑠 ≤ 1

Traditional (0 or 1)

MTTF MTTR

totalTotal

Capacity=1 Capacity=0

Different Network Failure Scenarios

Four types of network failures:

Fiber link failure

Microwave link failure

BS/ONU-BS node failure

OLT failure

16

PN

PN

FL

FLML

WOBAN

ML

BS node failure

OLT failure

Microwave link failure

Fiber link failure

𝐶𝑝𝑎𝑟𝑡𝑖𝑎𝑙𝑠

𝐶𝑝𝑎𝑟𝑡𝑖𝑎𝑙𝑠 = 𝑙∈𝑳

𝜆𝑙∙𝑑𝑙∙𝐵𝑝𝑎𝑟𝑡𝑖𝑎𝑙𝑠,𝑙

𝑊+ 𝑚∈𝑴

𝜆𝑚∙𝑑𝑚∙𝐵𝑝𝑎𝑟𝑡𝑖𝑎𝑙𝑠,𝑚

𝑊+ 𝑥∈𝑵

𝜆𝑥∙𝐵𝑝𝑎𝑟𝑡𝑖𝑎𝑙𝑠,𝑥

𝑊+

𝑝∈𝑷𝜆𝑝∙𝐵𝑝𝑎𝑟𝑡𝑖𝑎𝑙

𝑠,𝑝

𝑊∀𝑠 ∈ 𝑺

𝜆𝑙∙𝑑𝑙

𝑊: Failure rate of fiber link l relative to the entire physical network

𝐵𝑝𝑎𝑟𝑡𝑖𝑎𝑙𝑠,𝑙 : Total remaining capacity of slice s weighted by the mean time to

repair the failure of fiber link l that affects the slice

W: Mean failure rate of a WOBAN (four network failure scenarios)

𝑊 = 𝑙∈𝑳(𝜆𝑙 ∙ 𝑑𝑙) + 𝑚∈𝑴(𝜆𝑚 ∙ 𝑑𝑚) + 𝑥∈𝑵 𝜆𝑥 + 𝑝∈𝑷 𝜆𝑝

17

OLT failure

BS node failure

Fiber link failure Microwave link failure

Outline

Background

New Definition for Availability

Our Research Problem

Heuristic Algorithms

Test Conditions

Simulations and Performance Analyses

Conclusions

18

Research Problem

Objective:

Maximize: # of slices provisioned with guaranteed availability

𝑠∈𝑺 𝛿𝑠

Constraints:

Communication limitation:

Limited transmission capacity of each PON system

Limited transmission capacity of each microwave link

MEC limitation: Limited C/S capacity of each node (BS/ONU-BS)

Slice availability requirement: >= 0.99999

We formulate it as an Integer Linear Programing

(ILP) model!

19

Heuristic Algorithm

Key Steps

20

VN

VN

PN

PN

VL

FL

FLML

Slice

WOBAN

ML

VN

VN

PN

PN

VL

FL

FLML

Slice

WOBAN

ML

K1

K2

v1

Step 1: Virtual Node Mapping

Each virtual node is mapped to a

physical node

Judge whether the remaining C/S

capacity of each mapped physical

node is sufficient to satisfy the

demand of the virtual node

Each virtual link is mapped to one

of physical paths

According to different link cost

metrics, employ the shortest path

algorithm to establish the virtual

link along physical links with

sufficient remaining capacity

Step 2: Virtual Link Mapping

Link Cost Metric Definitions

Heu_Length algorithm

The link metric considers the load of each physical link in addition to its unavailability

The algorithm simultaneously balances the network traffic load and maximizes the slice availability

𝑐𝑙 = 𝑈𝑙 ∙ 𝑢𝑙

21

The link metric is based on the

length of each physical link

The physical length of a link

essentially corresponds to the

unavailability of the link since

they hold a linear relationship

Heu_Load algorithm

Link Cost Metric of Heu_LoadAlgorithm

1.𝑼𝒍 = 𝟏 − 𝑨𝒔 ∙ 𝑨𝒍

2.𝒖𝒍 =𝒄𝒖𝒔𝒆𝒅𝒄𝒕𝒐𝒕𝒂𝒍

3. 𝒄𝒍 = 𝑼𝒍 ∙ 𝒖𝒍

𝒍 𝒔 𝒅

Link l Link l+1 Link l+2

𝑈𝑙: The unavailability of the physical link l.

𝐴𝑠: The availability of the source node s.

𝐴𝑙: The availability of the physical link l.

𝑢𝑙 : The capacity utilization of the physical link l.

𝑐𝑢𝑠𝑒𝑑 : The used capacity of the physical link l.

𝑐𝑡𝑜𝑡𝑎𝑙 : The total capacity of the physical link l.

𝑐𝑙 : The cost of the physical link l.

22

Outline

Background

New Definition for Availability

Our Research Problem

Heuristic Algorithms

Test Conditions

Simulations and Performance Analyses

Conclusions

23

Test Conditions

Test case A

2 PONs, 10 ONU-BSs, 2 BSs, 22 microwave links

24

OLT

ONU-BS

ONU-BS

ONU-BS

OLT: Optical Line Terminal

RN: Remote Node

ONU-BS: Optical Network Unit-

Base Station

UE: User Equipment

ONU

VM

BS

Fiber Link

Microwave Link

ONU-BS

RN

RN

OLT

UE

BS

BS

ONU-BS

ONU-BS

ONU-BS

ONU-BS

ONU-BS

ONU-BS

Test Conditions

Test case B

5 PONs, 66 ONU-BSs, 9 BSs, GIS map

25

PON1

PON5PON2

PON4PON3

OLT

ONU-BS

BS

Fiber Link

Microwave Link

Test Conditions

26

Physical resources

Network slices

Maximum transmission capacity of each PON 10 Gb/s

Maximum transmission distance of a microwave link 20 km

Transmission capacity of a

microwave link

distance (d) < 10 km 3 Gb/s

10<distance (d)<20 km 3.6-0.06*d

C/S capacity at each physical node 100 VMs

Number of virtual nodes in a slice4 to N/2 (N is the total number of

physical nodes)

Number of virtual links in a slice V to 1.5V (V is the number of virtual

nodes in the slice)

Bandwidth of each virtual link 100 to 150 Mb/s

C/S capacity at each virtual node 4 VMs

# of Provisioned Slices

Heu_Load

Heu_Length

15

17

19

21

23

25

27

29

31

33

25 30 35 40 45 50 55

Nu

mb

er

of

Pro

vis

ion

ed

Slices

Number of request slices

ILP_Model

27

Limited communication resources affect the number of slices provisioned

Heu_Load algorithm performs closer to the ILP model

Heu_Load algorithm can provision more slices than Heu_Length algorithm

Case A

Impact of Node VM Resources

28

C/S capacity at each physical node affects the number of slices provisioned

Heu_Load algorithm is efficient to perform close to the ILP model and

outperforms Heu_Length algorithm

10

15

20

25

30

35

20 30 40 50 60 70

Nu

mb

er

of

Pro

vis

ion

ed

Slices

Number of VMs per physical node

ILP_Model

Heu_Load

Heu_Length

Case A

# of Provisioned Slices

Limited communication resources affect the number of slices provisioned

Heu_Load algorithm is more efficient to outperform Heu_Length algorithm

to provision more availability-guaranteed slices

25

27

29

31

33

35

37

39

41

43

30 35 40 45 50 55

Nu

mb

er

of

Pro

vis

ion

ed

Slices

Number of request slices

Heu_Load

Heu_Length

Case B

Impact of Node VM Resources

30

C/S capacity at each physical node affects the number of slices provisioned

Heu_Load algorithm is more efficient to outperform Heu_Length algorithm

20

25

30

35

40

45

40 50 60 70 80 90

Nu

mb

er

of

Pro

vis

ion

ed

Slices

Number of VMs per physical node

Heu_Load

Heu_Length

Case B

Conclusion

We maximize the number of availability-guaranteed

slices provisioned in a sliceable WOBAN that supports

MEC

By considering different network failure scenarios, a new

definition of availability is made for a slice that partially

functions

We formulate the slice provisioning problem using an

ILP model, and also develop two heuristic algorithms

based on different link cost metrics

Heu_Load algorithm is efficient to perform close to the

ILP model and outperforms Heu_Length algorithm

31

32

Thank you ！

Further Thinking

In the proposed scheme, all the network slices

are assumed to be static. However, the slice

services in a real network can often be

dynamical, so dynamic slice provisioning with

guaranteed availability can be further explored

33