qox oriented management mechanisms in towards-5g cloud ......end-to-end (e2e) latency,...

Fidel Liberal, Jose Oscar Fajardo, Bego Blanco, Rubén Solozabal, Ianire Taboada

[email protected]

QoX oriented management mechanisms in towards-5G cloud enabled RANs

Outline •  Introduction •  Towards-5G Cloud Enabled RANs (CRAN, MEC & FS) •  NFV & Optimization loops •  Example 1. C-RAN, MEC: Impact into higher layers •  Example 2. MEC: Channel-aware Service Level scheduling •  Example 3. Robust optimization on VNF placement

2

Disclaimer AKA excusatio non petita accusatio manifesta

5G Requirements User experienced

data rate(Mbit/s)

100

Spectrumefficiency

IMT-2020

3´

500

1106

10

20

100´Mobility(km/h)

Latency(ms)

Connection density(devices/km )

2

Networkenergy efficiency

Area trafficcapacity

(Mbit/s/m )2

Peak data rate(Gbit/s)

10

1´

400350

10105

10´1´

10.1

1

IMT-advanced

Source: ITU-R M.2083-0

Source: Face in Hole

Where’s Wally Pareto?

Multiple trade-off!!!

3/24

5G Requirements II

Source: 5G empowering vertical industries (5G-PPP)

AUTOMOTIVEDATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY

Automated drivingShare my viewBird’s eye viewDigitalization of transport and logisticsInformation society on the road

e-HEALTH

Assets and interventions management in HospitalRoboticsRemote monitoringSmarter medication

DATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY

Grid backhaulGrid access

Grid backbone

ENERGYDATA RATE

COVERAGEMOBILITY(SPEED)

(LOW)LATENCY

POSITIONINGACCURACY

Collaborative gaming

Ultra high fi delity mediaOn-site Live Event Experience User/Machine generated contentImmersive and integrated mediaCooperative media production

MEDIA & ENTERTAINMENT

DATA RATE

DENSITYRELIABILITY


(LOW)LATENCY

POSITIONINGACCURACY

Additional Capabilities

• Service Deployment Time.• Data Volume.• Autonomy.• Security.• Identity.

FIGURE 2: VERTICALS SECTORS’ CAPABILITIES AND REQUIREMENTS

SPIDERS CHARTS

DENSITY

RELIABILITY

1110

and eHealth with maximum value in the order of 500 km/h..

• E2E Latency: Maximum tolerable elapsed time from the instant a data packet is generated

at the source application to the instant it is received by the destination application. If direct

mode is used, this is essentially the maximum tolerable air interface latency. If infrastructure

mode is used, this includes the time needed for uplink, any necessary routing in the

infrastructure, and downlink. The most demanding vertical use cases are related to Factories

with minimum values of 100 µs to 10 ms.

• Density (number of devices): Maximum number of devices (vehicles in the case of Automotive)

per unit area that are 5G capable, although they might not all be generating traffi c simultaneously for the specifi ed application. The most demanding vertical use cases are related to Factories with up to 100/m2.

• Reliability: Maximum tolerable packet loss rate at the application layer within the maximum

tolerable end-to-end latency for that application. The most demanding vertical use cases are

related to eHealth with values up to 99.99999%.

• Position Accuracy (Location): Maximum positioning error tolerated by the application. The

most demanding vertical use cases are related to Automotive with minimum values in the order

of 0.3 m.

In addition to these ITU-like capabilities, the Coverage capability is also assessed as critical for the

diff erent vertical sectors: • Coverage: Area within which or population for which the application should function

correctly, i.e. the specifi ed requirements (latency, reliability and data rate) are achieved. Most of the vertical sectors have strong requirements on geographic and/or population coverage.

The analysis of the key requirements from the diff erent vertical sectors on this critical baseline lead to the spider charts captured in Figure 2. The quantifi cation is based on the following ranking: (0) No requirement, (1) Low level of requirement or no specifi c constraint, (2) Medium level of requirement, which could be satisfi ed with existing legacy systems (3) High level of requirement, which may be at the limit or not satisfi ed with the existing legacy systems and (4) Very high level of requirement, corresponding to the 5G Infrastructure PPP targets and Key Performance Indicators (KPIs).

No single vertical use case requires the full set of 5G capabilities to be met at the same time. The

5G system will be relying on a dynamic and fl exible function allocation and confi guration and 5G networks will highly rely on software networking, virtualization and slicing techniques. It is expected

that the softwarisation of infrastructure composition and usage will allow addressing various use

cases from the vertical sectors, as detailed in Chapter 4.

The following additional capabilities are also assessed as key from the vertical sectors perspectives:

• Service Deployment Time: Duration required for setting up end-to-end logical network

slices characterised by respective network level guarantees (such as bandwidth guarantees,

End-to-End (E2E) latency, reliability…) required for supporting services of that particular vertical

sector. Programmable networks and multi-tenant capability in 5G will ensure speedy deployment

of services (e.g. 5G Infrastructure PPP targets 90 minutes for service deployment).

• Data Volume: Quantity of information transferred (downlink and uplink) per time interval

over a dedicated area (e.g. 5G Infrastructure PPP targets a maximum of 10 Tb/s/km2).

• Autonomy: Time duration for a component to be operational without power being supplied.

It relates to battery lifetime, battery load capacity and energy effi ciency. • Security: System characteristic ensuring globally the protection of resources and

encompassing several dimensions such as authentication, data confi dentiality, data integrity, access control, non-repudiation…

• Identity: Characteristic to identify sources of content and recognise entities in the system.

One key parameter to guarantee the fast adoption of 5G is the possibility to access low cost

solutions in several use cases of the vertical sectors.

5G VERTICAL SECTORS: BEYOND TODAY’S NETWORK CAPABILITIESThrough an integration of various radio access technologies and Device-to-Device (D2D) commu-

nication, 5G is expected to provide the coverage needed to support road safety applications eve-

rywhere. Additionally, by targeting an end-to-end latency of 5 ms (down to 1 ms for direct mode)

with extreme network reliability and enabling scalability of solutions by providing deterministic per-

formances also at high load, 5G is envisioned to be a key enabler for automated driving and related

critical services for which the stringent requirements could not yet met by existing technologies.

Manufacturing is one of the most demanding industry in terms of 5G networking support (mobility

and wireless broadband), requiring ultra-high reliability, latencies down to 1 ms (for real time

process control), and densities of more than 10 to 100 machine sensor streams per square meter.

Existing legacy technologies however do not handle mobility, especially in terms of handovers. In

addition, the bandwidth per object is very limited (100 b/s, < 1ko/day) and the latency is high (one

full second may be needed to transmit a message), thus not allowing the deployment of real time

applications. Also, due to the utilization of unlicensed bands, these technologies lack dedicated

quality-of-service guarantees. ■

The following chapter reviews the target performance parameters for 5G, currently being assessed globally and discuss the key performance parameters derived from a comprehensive analysis of the use cases

introduced in Chapter 1, as well as present a mapping of the most plausible vertical scenarios onto the new fundamental capabilities of 5G.

5G KEY CAPABILITIES AND KPIsAs described in the 5G Infrastructure PPP Vision (March 201516), the 5G capabilities will

provide ubiquitous access to a wide range of applications and services with increased resilience,

continuity, and much higher resource effi ciency, while protecting security and privacy. In addition, 5G will provide enormous improvements in capacity and boost user data rates. The

highly demanding capabilities of 5G require an outstanding research and innovation eff ort to reach orders of magnitude of improvement over the current technology and infrastructure.

The following 5G targets, which are under further discussion worldwide, in bodies such

as International Telecommunication Union-Radiocommunication Sector (ITU-R), 3rd

Generation Partnership Project (3GPP), and Next Generation Mobile Networks (NGMN)

Alliance, indicate the advances of 5G systems compared to previous generation:

• 1,000 X in mobile data volume per geographical area reaching a target of 0.75 Tb/s

for a stadium.

• 1,000 X in number of connected devices reaching a density ≥ 1M terminals/km2.

• 100 X in user data rate reaching a peak terminal data rate ≥1 Gb/s for cloud

applications inside offi ces. • 1/10 X in energy consumption compared to 2010 while traffi c is increasing dramatically at the same time.

• 1/5 X in end-to-end latency reaching delays ≤ 5 ms.

• 1/5 X in network management Operational Expenditure (OPEX).

• 1/1,000 X in service deployment time reaching a complete deployment in

≤ 90 minutes.

• Guaranteed user data rate ≥ 50 Mb/s.

• Capable of IoT terminals ≥ 1 trillion.

• Service reliability ≥ 99.999% for specifi c mission critical services. • Mobility support at speed ≥ 500 km/h for ground transportation.

• Accuracy of outdoor terminal location ≤ 1 m.

IDENTIFYING 5G VERTICAL SECTORS USE CASES AND REQUIREMENTSThere is a worldwide eff ort on the further characterization of 5G use cases and related requirements (e.g. ITU, 3GPP SMARTER, NGMN...). The 5G Infrastructure Public Private

Partnership (PPP) is progressing on the identifi cation of uses cases with a clear eff ort on verticals markets, e.g. the fi ve white papers addressing Factories, Automotive, eHealth, En-ergy as well as Media and Entertainment.

Considering the 5G Infrastructure Association vertical use cases introduced in Chapter 1,

the three ITU-R usage scenarios17 (Enhanced Mobile Broadband, Ultra-reliable and Low

Latency Communications and Massive Machine Type Communications and the related

spider diagram capabilities), the eight NGMN use case families18 and the four 3GPP use

cases groups19 (Enhanced Mobile Broadband (eMBB), Massive IoT (mIoT), Critical

Communications (CriC) and Network Operation (NEO)) covering more than seventy distinct

use cases, this section presents the 5G Infrastructure PPP vertical use cases capabilities

spiders, considering the major relevant capabilities of each vertical sector.

Several ITU-like capabilities can be considered as critical parameters for the diff erent vertical sectors:

• Data Rate: Required bit rate for the application to function correctly. It corresponds

to the user experienced data rate as defi ned by ITU. The most demanding vertical use cases are related to Media & Entertainment with maximum values in the order of Gb/s.

• Mobility (speed): Maximum relative speed under which the specifi ed reliability should be achieved. The most demanding vertical use cases are related to Automotive

TECHNICAL REQUIREMENTS 3

u http://5g-ppp.eu/wp-content/uploads/2015/02/5G-Vision-Brochure-v1.pdf

i http://www.itu.int/dms_pubrec/itu-r/rec/m/R-REC-M.2083-0-201509-I!!PDF-E.pdf

o https://www.ngmn.org/uploads/media/NGMN_5G_White_Paper_V1_0.pdf

p http://www.3gpp.org/ftp/Specs/archive/22_series/22.891/22891-110.zip

FACTORIES

Time-critical process controlNon time-critical factory automationRemote controlIntra/Inter-enterprise communicationConnected goods

DATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY

AUTOMOTIVEDATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY


e-HEALTH


DATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY


Grid backbone

ENERGYDATA RATE


(LOW)LATENCY

POSITIONINGACCURACY




DATA RATE

DENSITYRELIABILITY


(LOW)LATENCY

POSITIONINGACCURACY




SPIDERS CHARTS

DENSITY

RELIABILITY

1110















of 0.3 m.
















































for a stadium.







≤ 90 minutes.
























FACTORIES


DATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY

AUTOMOTIVEDATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY


e-HEALTH


DATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY


Grid backbone

ENERGYDATA RATE


(LOW)LATENCY

POSITIONINGACCURACY




DATA RATE

DENSITYRELIABILITY


(LOW)LATENCY

POSITIONINGACCURACY




SPIDERS CHARTS

DENSITY

RELIABILITY

1110















of 0.3 m.
















































for a stadium.







≤ 90 minutes.
























FACTORIES


DATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY

AUTOMOTIVEDATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY


e-HEALTH


DATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY


Grid backbone

ENERGYDATA RATE


(LOW)LATENCY

POSITIONINGACCURACY




DATA RATE

DENSITYRELIABILITY


(LOW)LATENCY

POSITIONINGACCURACY




SPIDERS CHARTS

DENSITY

RELIABILITY

1110















of 0.3 m.
















































for a stadium.







≤ 90 minutes.
























FACTORIES


DATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY

AUTOMOTIVEDATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY


e-HEALTH


DATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY


Grid backbone

ENERGYDATA RATE


(LOW)LATENCY

POSITIONINGACCURACY




DATA RATE

DENSITYRELIABILITY


(LOW)LATENCY

POSITIONINGACCURACY




SPIDERS CHARTS

DENSITY

RELIABILITY

1110















of 0.3 m.
















































for a stadium.







≤ 90 minutes.
























FACTORIES


DATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY

AUTOMOTIVEDATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY


e-HEALTH


DATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY


Grid backbone

ENERGYDATA RATE


(LOW)LATENCY

POSITIONINGACCURACY




DATA RATE

DENSITYRELIABILITY


(LOW)LATENCY

POSITIONINGACCURACY




SPIDERS CHARTS

DENSITY

RELIABILITY

1110















of 0.3 m.
















































for a stadium.







≤ 90 minutes.
























FACTORIES


DATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY

Closed?

4/24

5G Requirements: Beyond NR

5G challenges: -  Low Latency services - Hi Bandwitdh- Ultra-dense - HighTraffic (IoT) - Mission-Critical services

- CAPital EXpenditure (CAPEX) - OPerational EXpenditures

(OPEX) VS

5G technology pillars:

-  Aggregation of HW resources Virtualization: SDN & NFV & MEC

5/24 Ref: Technology pillars in the architecture of future 5G mobile networks: NFV, MEC and SDN, “Blanco et al.

LTE

Single eNB

UEeNodeB

Split of digital and radio parts

UERRHBBU

eNodeB

Mobile fronthaul

Moving towards 5G => CE-RAN

VNFs PNFs

Ref: Small Cell Forum 6/24

4G à 5G

v-BBU

v-BBU

v-BBU

RRHRRH

RRH

RRH

RRH

UE

Macro-cells

Small cells

CloudSDR-based

BBU

NFV-based

fronthaul

v-fronthaul

hetnet densification

C-RAN vs. MEC

RRM

Mobile fronthaul

...

Centralized functions

...

Mobile fronthaul

MEC

...

MEC

Remote functions

Centralized functions

Remote functions

a) fully centralized RAN b) partially centralized RAN

...RRM

7/24

MEC I

Centralized CE-RAN

PNF

MEC II

MAC

RLC

PHY

RRC PDCP

SON OAM APP

RF

PNFs

VNFs

Functional split III

MAC

RLC

PHY

RRC PDCP

PNFs

SON OAM APP

VNFs

Functional split II

Functional split I

APP

VNFs

SC

Distributed CE-RAN

PNF

SC

PNF

SC

Hybrid MANO

RRH

BBU

Fully centralized CE-RAN Distributed CE-RAN

EPC

5G Cloud enabled Radio Access Network

8/24

5G NFV (ETSI MANO)

9/24

Optimization loops VIM

-VNFM

LightD

C

OBSERVE PLAN

DECIDEACT

LEARN

NVFO

VNFM

-VIM

OBSERVE PLAN

DECIDEACT

LEARN

NMS-EM

SNFV

O+V

NF

OBSERVE PLAN

DECIDEACT

LEARN

VNFVN

F

OBSERVE PLAN

DECIDEACT

LEARN

10/24

Optimization loops II options

•  Technologically improviing/optimizing a specific VNF itself => example 1 and 2 –  Due to reduced latencies, MEC etc –  Tradeoff delay and effect into

•  TP and HARQ => Example 1 •  Channel-aware scheduling => Example 2

•  Optimizing the overall system => example 3 –  Multi-tenant VNF placement

11/24

Impact of C-RAN on higher layers

•  COST Across TF6 related –  Impact of reducing delay into transport protocols

•  Mobility Requirement!!!!

–  MEC or 5G scenarios (C-RANs)

•  HARQ (UPV/EHU with UMA) –  Functional split vs. Non ideal fronthaul?

Example 1

12/24

Impact of C-RAN on higher layers •  Impact of reducing delay

–  MEC or 5G scenarios (Transport end-points closer –delay-)

Example 1

Context

Mobilityscenario1under4Glatencies

Mobilitypattern1 Mobilitypattern2

Mean MeanCI Mean MeanCI

CUBIC 24.18 6.59 21.87 6.89

NewReno 22.95 7.95 19.87 7.16

BBR 25.43 18.7 24.42 17.54

Westwood+ 16.08 8.01 17.16 7.68

Illinois 26.2 10.7 23.35 14.84

Context

Mobilityscenario2under4Glatencies



CUBIC 12.62 7.81 8.76 5.51

NewReno 10.63 7.58 4.15 2.89

BBR 9.09 10.21 9 6.76

Westwood+ 8.61 4.13 7.58 4.28

Illinois 11.48 6.8 8.95 5.84

Context

Mobilityscenario1underlowlatencies



CUBIC 28.18 8.84 28.21 9.02

NewReno 28.68 9.4 24.89 10.36

BBR 28.69 8.29 26.84 8.41

Westwood+ 16.57 10.83 20.3 12.27

Illinois 28.19 9.1 23.81 11.78

Context

Mobilityscenario2underlowlatencies



CUBIC 10.33 5.98 11.91 5.4

NewReno 15.79 5.86 12.43 4.97

BBR 14.66 5.38 9.35 5.32

Westwood+ 14.66 4.76 8.11 5.58

Illinois 15.82 5.48 11.46 4.92

13/24

Impact of C-RAN on higher layers

•  HARQ (UPV/EHU ongoing currently with UMA) –  Hybrid & non ideal fronthauls

•  Effect of HARQ/NACK and CQI report delaying

Example 1

14/24 Ref: Impact of Front-Haul Delays in Non-Ideal Cloud Radio Access Networks” Carreras et al.

MEC as channel-aware scheduling enabler

•  Optimization idea: –  Use radio channel state to optimize radio transmission:

•  Radio (L2) level •  Service level

•  Scenario –  Towards 5G network

•  Option a) –  Full C-RAN

•  Option b) –  MEC

Example 2

15/24

Ref: “Radio-aware service-level scheduling to minimize downlink traffic delay through MEC” Fajardo et al.

Example: MEC 2-layer traffic scheduler

Cloud-enabledRadio Access Network

Mobile Backhaul

Core Network

Traffic sources

Mobile EdgeScheduler

Incoming traffic flows Flow- and

channel-aware scheduling

Cell load

Fine-grain CQI reporting

(CRR)

CRR = CQI Reporting Rate to eNodeBCG-CRR = CQI Reporting Rate to MEC scheduler

eNodeB

MECServer

Coarse-grain CQI reporting

(CG-CRR)

Shaped traffic flows offered to eNodeB

Traffic flows scheduled by

eNodeB

(1) (2)

(1) (2)

0 0.5 1 1.5 2 2.5

x 105

0

5

10

15

CQI

Time (ms)

CQI eNodeB (5ms)CQI MESch (1s)

250 500 750 1000 1250 1500 1750 2000 2250 2500 2750 3000 3250 3500 3750 4000-4000

-3000

-2000

-1000

0

1000

2000

3000

4000

5000

CG-CRR (ms)

Est

imat

ed T

hr. -

Act

ual T

hr. (

kbps

)

Average Thr. OverestimationAverage Thr. Underestimation

250 500 750 1000 1250 1500 1750 2000 2250 2500 2750 3000 3250 3500 3750 40000

200

400

600

800

1000

1200

1400

CG-CRR (ms)

Ave

rage

Thr

. Mis

s-es

timat

ion

(kbp

s)

EVALD Inst. CQI

EVALD Median CQI

EVAHD Inst. CQI

EVAHD Median CQI

0.3 0.4 0.5 0.6 0.7 0.8 0.90

10

20

30

40

50

60

ρ

Mea

n de

lay

(s)

eNodeB(BC)MESch-eNodeB(BC)eNodeB(MASPI)

16/24

Example: MEC multimedia scheduler

Cloud-enabledRadio Access Network

Mobile core

network

Core Network

DASHServer

DASHClient

ME-DAFDASH SVC

cache

HTTP GETHTTP GET

(new segments)DASH

Manager

Adaptation Algorithm

RAN monitor

Cell load

Link quality

Fine-grain CQI reporting

(CRR)

CRR = CQI Reporting Rate to eNodeBCRR-2 = CQI Reporting Rate to ME-DAF

eNodeB

MECServer

Coarse-grain CQI reporting

(CRR-2)

0 0.5 1 1.5 2 2.5

x 105

0

5

10

15

CQI

Time (ms)

Class-1 CQI eNodeBClass-2 CQI eNodeBClass-1 CQI ME-DAFClass-2 CQI ME-DAF

250 500 750 1000 1250 1500 1750 2000 2250 2500 2750 3000 3250 3500 3750 4000-1500

-1000

-500

0

500

1000

1500

2000

CG-CRR (ms)

Est

imat

ed T

hr. -

Act

ual T

hr. (

kbps

)

Average Thr. OverestimationAverage Thr. Underestimation

250 500 750 1000 1250 1500 1750 2000 2250 2500 2750 3000 3250 3500 3750 40000

200

400

600

800

1000

1200

1400

CG-CRR (ms)

Ave

rage

Thr

. Mis

s-es

timat

ion

(kbp

s)

"Class-1" Inst. CQI"Class-1" Median CQI"Class-2" Inst. CQI"Class-2" Median CQI

eNodeB Client C-MEC CB-MEC1

1.5

2

2.5

3

3.5

4

4.5

5BC

MO

S

AggregationClass-1 UEsClass-2 UEs

eNodeB Client C-MEC CB-MEC1

1.5

2

2.5

3

3.5

4

4.5

5LC

MO

S

AggregationClass-1 UEsClass-2 UEs

(b)(a)

17/24 Ref: “Improving Content Delivery Efficiency through Multi-Layer Mobile Edge Adaptation ” Fajardo et al.

Service Level Agreement (Multiple Tenants deployment in a Football Stadium)

Challenge: “ensuring QoS per tenant” while minimizing power consumption

Robust Optimization and VNF placement Example 3 VIRT

UAL

INFRAS

TRUCT

URE

CORENETWORK

SCVNF1

SCVNF1

SCVNF1

SCVNF1 SCVNF1

SCVNF1

SCVNF1

SCVNF1

SCVNF1

SCVNF1

SCVNF1

SCVNF SCVNF

vFW

VA

vFW

vID*

vWM

vCachevFW

vCache

vTU

SCVNF1

SCVNF1 SCVNF1

SCVNF1

SCVNF1

SCVNF1

SCVNF1

SCVNF1

SCVNF1

HWAcc

CPUs

vID

NS2

NS1

NS1

NS1

NS1

NS1

NS1

NS1 N

S3

NS3N

S1

NS2

Networktopology

vFW vCache vTU

vFW VA

vFW vID vWM vCache <lat1

<lat2

<lat3

TENAN

T1

SCVNF

SCVNF

TENAN

T2

NS1

NS2

NS3

MEC I

Centralized CE-RAN

PNF

MEC II

MAC

RLC

PHY

RRC PDCP

SON OAM APP

RF

PNFs

VNFs

Functional split III

MAC

RLC

PHY

RRC PDCP

PNFs

SON OAM APP

VNFs

Functional split II

Functional split I

APP

VNFs

SC

Distributed CE-RAN

PNF

SC

PNF

SC

Hybrid MANO

RRH

BBU

Fully centralized CE-RAN Distributed CE-RAN

EPC

18/24

AUTOMOTIVEDATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY


e-HEALTH


DATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY


Grid backbone

ENERGYDATA RATE


(LOW)LATENCY

POSITIONINGACCURACY




DATA RATE

DENSITYRELIABILITY


(LOW)LATENCY

POSITIONINGACCURACY




SPIDERS CHARTS

DENSITY

RELIABILITY

1110















of 0.3 m.
















































for a stadium.







≤ 90 minutes.
























FACTORIES


DATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY

AUTOMOTIVEDATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY


e-HEALTH


DATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY


Grid backbone

ENERGYDATA RATE


(LOW)LATENCY

POSITIONINGACCURACY




DATA RATE

DENSITYRELIABILITY


(LOW)LATENCY

POSITIONINGACCURACY




SPIDERS CHARTS

DENSITY

RELIABILITY

1110















of 0.3 m.
















































for a stadium.







≤ 90 minutes.
























FACTORIES


DATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY

AUTOMOTIVEDATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY


e-HEALTH


DATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY


Grid backbone

ENERGYDATA RATE


(LOW)LATENCY

POSITIONINGACCURACY




DATA RATE

DENSITYRELIABILITY


(LOW)LATENCY

POSITIONINGACCURACY




SPIDERS CHARTS

DENSITY

RELIABILITY

1110















of 0.3 m.
















































for a stadium.







≤ 90 minutes.
























FACTORIES


DATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY

AUTOMOTIVEDATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY


e-HEALTH


DATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY


Grid backbone

ENERGYDATA RATE


(LOW)LATENCY

POSITIONINGACCURACY




DATA RATE

DENSITYRELIABILITY


(LOW)LATENCY

POSITIONINGACCURACY




SPIDERS CHARTS

DENSITY

RELIABILITY

1110















of 0.3 m.
















































for a stadium.







≤ 90 minutes.
























FACTORIES


DATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY

AUTOMOTIVEDATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY


e-HEALTH


DATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY


Grid backbone

ENERGYDATA RATE


(LOW)LATENCY

POSITIONINGACCURACY




DATA RATE

DENSITYRELIABILITY


(LOW)LATENCY

POSITIONINGACCURACY




SPIDERS CHARTS

DENSITY

RELIABILITY

1110















of 0.3 m.
















































for a stadium.







≤ 90 minutes.
























FACTORIES


DATA RATE

DENSITYRELIABILITY

COVERAGEMOBILITY

(LOW)POSITIONING

(SPEED)

LATENCYACCURACY

Ref:A Robust Optimization Based Energy-Aware Virtual Network Function Placement Proposal for Small Cell 5G Networks with Mobile Edge Computing Capabilities Blanco et al.

VNF1

VNF1

VNF1 VNF2 VNF2

VNF2 VNF3 VNF3 VNF5

VNF1 VNF3 VNF4 VNF4 VNF6

VNF2 VNF3 VNF4

VNF1 VNF2 VNF3

VNF1 VNF2 VNF3 VNF4 VNF5 VNF6

VNF1

VNF1 VNF2 VNF2

VNF2 VNF3 VNF3 VNF5

VNF1 VNF3 VNF4 VNF4 VNF6

< lat1

< lat2

< lat3

RAM

HD

CPU

Network topology Link

capacity and delay

Proposed solution:

Discrete combinational optimization problem

Constraint programming

Robust optimization techniques (variable traffic demand)

FlatZinc with GeCode constraint solver

http://www.gecode.org/flatzinc.html

Robust Optimization and VNF placement Example 3

Stochastic modeling of the whole system => complex multi-dimensional coupled problem => Dead End => A. Kassler (KAU )+ RO

19/24

Model constraints:

Network service:

Latency of a Network Service

Global Power Consumption

•  SCVNF processing time•  Service VNF processing time•  Network transmission time

•  Power consumption of CPUs•  Power consumption of cell chassis•  Power consumption of switch

Service flavor:

Cost function:

Aggregated Traffic •  Nominal traffic + Robust protection


20/24

SC Modeling:

• VNFs are modeled to use resources (i.e. %CPU, RAM ...) according to the traffic load (aggregated bitrate served)

• Number of VM needed depend on the aggregated traffic• Latency introduced by the VNF is modeled using the aggregated traffic relative to

the total link capacity available• HW accelerators compatible VNFs allow to have a reduction in the latency and

consumption of the VNF


21/24

Network Modeling:

• Variable link latency with the load of the link and the size of the flow.• Power consumption of the network elements is proportional to the percentage of usage of

the link capacity.• Switching delay is proportional to the aggregated traffic related to the total link capacity.


22/24

NS placement:

NS service request:


23/24

Conclusions

•  Towards 5G –  NFV/SDN, CUPS, whatever…

•  Transition between 4G => 5G •  Multiple optimization posibilities

–  Complex ecosystem –  Multiple criteria –  Centralised vs. Distributed

•  CRAN vs. SON

24/24

Fidel Liberal, Jose Oscar Fajardo, Bego Blanco, Rubén Solozabal, Ianire Taboada

[email protected]

QoX oriented management mechanisms in towards-5G cloud enabled RANs

qox oriented management mechanisms in towards-5g cloud ......end-to-end (e2e) latency,...

Documents