foster univ - etsi · pdf filequality for best physical layer rf throughput. commercial...

ETSI SUMMITON5G NETWORK INFRASTRUCTURE

5G NEXT GENERATION PROTOCOLS – LAST CHANCE TO CHANGE

Presented by Gerry Foster

© All rights reserved

ETSI Summit5G Network Infrastructure2 © All rights reserved

Mobile Internet is one of the most successful technical achievements of our time.

ETSI Summit5G Network Infrastructure

… its success is largely due to key SDOs involved

3 © All rights reserved

3GPP cellular is the most successful mobile system today– by global roaming mobility, revenue & # citizens connected

GSM is the most efficient mobile voice system today

GPRS, UMTS and LTE have added mobile internet – National commercial statistics for many countries report that more sessions start & end on

mobiles than any other access technology.– mobile access is becoming the norm!

The IETF has provided the networking protocols to enable mobile internet

IP Networking

TCP Transmission

HTTP/ HTML internet content exchange & language

SCTP/DIAMETER Signalling: LTE‐S1‐Access & CORE signalling

SIP Media sessions used in VoLTE


So what’s the problem?


3GPP designs for mobile access connectivity & centralised subscriber control

4G-LTE provides a well received integrated ‘mobile Internet access’ technology

… however, in bringing these technologies together, users note:

Expensive subscriptions to get 3GPP & IP to work together, which need to be cheaper.

Variable performance (throughput & latency) c.f . fixed broadband & patchy coverage

The Internet has been designed for fixed networking and internetworking

ETSI Summit5G Network Infrastructure© All rights reserved

SO WHAT’S THE PROBLEM? … A CLO SER LOOK


3GPP LTE Protocol Architecture User Plane (UP)

6

App(user)

TCP(user)

IP(user)

PDCP

RLC

MAC

PHY

Bridge(PDCP<‐>GTP‐U)

Security Gateway

GTP‐U Bridge(S1‐S5)

GTP‐U(S5)

UDP(Nwk)

IP(Nwk)

Eth/Nwk

GTP‐U(S1)

UDP(Nwk)

IP(Nwk)

Eth/Nwk

GTP‐u(S5)

UDP(Nwk)

IP(Nwk)

Eth/Nwk

Gateway

App(user)

TCP(NAT)

IP(Nwk)

Eth/Nwk

GTP S1 Tunnel (S1-Bearer)

ESP Tunnel

GTP S5 Tunnel (S5-Bearer)

NAT translation

TCP(user)

IP(user)

PDCP

RLC

MAC

PHY

GTP‐U(S1)

UDP(Nwk)

IP(Nwk)

IP Sec (ESP)

Eth/Nwk

IP(Nwk)

IP Sec (ESP)

IP(Nwk)

Eth/Nwk

Eth/Nwk

NAT

IP(Nwk)

Eth/Nwk

TCP(NAT)

LTE-Uu S1-UP S5 SGi

RRC(DRB)

Note: ERAB = DRB + S1BearerEPS Bearer = ERAB + S5 BearerETE Service = EPS Bearer + External Bea

1.User IP is tunnelled from UE to SeGW/PGW using 2x GTP bearer tunnels

2.There is with no inherent User Plane (UP) security

3.There is no user level IP routing over cellular access

4.GTP tunnels have to be updated every mobility move

ETE UP has many protocol Bridges & Gateways (tunnel proc., tunnel header OHD’s, ROHC/ RAN‐IPsec. explicit signalling)

=> adds delay & processing cost


… & the Control Plane (CP)


UDP(Nwk)UDP(Nwk)UDP(Nwk)UDP(Nwk)

NAS(EMM)NAS(EMM)

PDCPPDCP

RLCRLC

MACMAC

PHYPHY

PDCPPDCP

RLCRLC

MACMAC

PHYPHY

AS/NAS Fn(RRC<‐>S1‐AP)AS/NAS Fn(RRC<‐>S1‐AP)

NAS Fn(EMM<‐>ESM) NAS Fn(EMM<‐>ESM)

SCTPSCTP

IP(Nwk)IP(Nwk)

Eth/NwkEth/Nwk

GTP‐C(S5)GTP‐C(S5)

UDP(Nwk)UDP(Nwk)

IP(Nwk)IP(Nwk)

Eth/NwkEth/Nwk

S1 UE ContextGTP S11 Control Session

GTP S5 Control Session

RRCRRC

SCTPSCTP

RRCRRC

NAS(EMM)NAS(EMM) NAS(ESM)NAS(ESM)

S1‐APS1‐APS1‐APS1‐AP

IP(Nwk)IP(Nwk)

Eth/NwkEth/Nwk

IP(Nwk)IP(Nwk)

Eth/NwkEth/Nwk

GTP‐C(S11)GTP‐C(S11)GTP‐C(S11)GTP‐C(S11)

IP(Nwk)IP(Nwk)

Eth/NwkEth/Nwk

UDP(Nwk)UDP(Nwk)

IP(Nwk)IP(Nwk)

Eth/NwkEth/Nwk

GTP‐C(S5)GTP‐C(S5)

GTP‐C Bridge(S11<‐>S5)GTP‐C Bridge(S11<‐>S5)LTE-Uu S1-MME S11 S5

RRC(SRB)

AS Security Binding

NAS(ESM)NAS(ESM)

NAS Security Binding

Security GatewaySecurity Gateway

IP(Nwk)IP(Nwk)

IPSec(ESP)IPSec(ESP)

IP(NwK)IP(NwK)

Eth/NwkEth/Nwk

Eth/NwkEth/NwkIPSec(ESP)IPSec(ESP)

ESP Tunnel

1. IP Packets can’t flow between Mobile & Internet until NAS/ EPS Control Plane (CP) bearer setup signalling is completed

2. EPS bearer setup needs a separate control protocol, GTP‐C

3. There is mandated fixed CP security (Authentication, Encryption Integrity) ETE CP bearer setup of the UP adds user experience latency

=> adds delay & processing cost


3GPP Mobile Internet Performance Improvement


KPI 1991(GSM)

Today(LTE)

Predicted for 5G (2020)

Mean Improvement

(from 1991 to today)Notes

Spectrum efficiency(bit/s)/Hz 0.45 16‐30 35 mean x35

Access Session Delay (ms) 3,000 100 10 x30 Radio Access Technology: RRC Connection & NAS Service

Request setup (typical case)

IP MTU Delay (ms) 400 30 1 x13 UE to edge of Cellular Network, user plane

RF Throughput(Mbit/s) 0.096 320 5,000 x5,333

Peak single user per Cell, in coverage with suitable RF quality for best Physical layer RF Throughput. Commercial LTE‐A. GSM was a basic circuit switched version. The predicted value for 5G is for a 3.5GHz RF channel of 80MHz bandwidth. Dense urban area.

TCP/IPThroughput(Mbit/s)

0.03 30 450 x1,000 Per RF single channel cell, dense urban cell.Mean per user with 5‐10 users per cell.

Med. Spectrum Efficiency gains expensive & now close to the limit

Med. Access delay gains, but Close to limit for subscription based

Mobile Access

Small gains in Latency … hardest KPI to improve

Notable 3GPP throughput gains needed for 3GPP+IP, ETE gain.

RF to TCP/IP User Level Thp increase ~ 5:1

In summary:

Mobile Internet as an ETE community has not put enough effort into making the Network part more efficient.

Most of the effort/cost has been invested in the Radio technology

This is why we all pay for the headline ‘radio throughput gains’ without realising that the Network technology enhancements are not keeping up

… whilst for 5G, the Radio technology is nearing its limits.


… and the macro scale security problem


There is no inherent user User‐Plane Network access security over the Mobile Internet!

SSL, TLS and HTTPS solve the wrong problem: • one size fits all security that is not needed for many applications and services• only provides encryption and is only ETE• adds overhead to all packets

Today users are much more discerning: • users are not ‘one size fits all’• Banking, IoT, Browsing all have different security needs, some at App level and some at network• All need basic user plane authentication of some kind but this can be Stakeholder peered and does not need to be

centralised all the time

Security Authority should not be a gateway but peered stakeholder security according to scope and resolution of the communication.

If we continue with no basic Stakeholder‐based Authentication for NwK UP access then security threat management will cost a large part of overall network OPEX in 5G.

We need Flexible, Scalable, Multi‐Homing network security

We have to chose between privacy and secure user plane access authentication beyond basic control plane subscription authentication.

ETSI Summit5G Network Infrastructure© All rights reserved

LAST CHANCE TO CHANGE?


Why Change for 5G?


If we don’t improve BS to Internet protocols for 5G, then radio investment will be wasted on legacy GTP & IP inefficiencies across AMPS: addressing, mobility, performance and scalability aspects.

Operators are sweating IP & GTP assets. However, there will be a tipping point when

• Sweat benefit << NGP efficiency & performance improvements (tipping point 5G ?)

Operators need new revenue potential

• justifies adding/ migrating to 5G to support AR/ VR, 4K‐TV & IoT.

• but these services are not efficiently supported over a RAN operating IP/ GTP

Better Mobile internet Access in next commercial 10 year timeframe needs definition of NGP at SDO level now!

It is time for 3GPP & IETF to standardise together rather than reference each other?


So what’s NGP all about?


NGP includes Vendors, Operators & Standards reps, and aims to:

• Provide Scenarios & Requirements for significantly enhanced 5G/ NGN, MOBILE INTERNET ACCESS protocol architecture

• Work with SDOs (3GPP, IETF, ITU‐T, IEEE) to stimulate NGP based MOBILE INTERNET ACCESS protocol(s) definition

IS NOT … trying to change the InternetIS … trying to significantly improve MOBILE INTERNET ACCESS SUPPORTS … existing IETF protocol evolution in the context of MOBILE INTERNET ACCESS

Industry view has been that IETF are best placed to develop an NGP solution:

•However, IETF 97 stated that IPv6 is the IETF preferred future networking standard•This decision does not meet NGP requirements for MOBILE INTERNET ACCESS •Therefore, 3GPP is the most likely candidate SDO to lead NGP realisation of better MOBILE INTERNET ACCESS NGPs … in sync with IETF, ITU‐T, & IEEE

NGP understands that in 2020 ‘MOBILE’ includes: Static‐Nomadic, RF‐Cellular, RF‐Millimetric & Wi‐Fi


NGP Requirement Summary


Efficiency Need much Smaller Headers over Access Technologies, for NGP packets to maximise access efficiency, in particular spectrum efficiency for radio technologies

Security Needs to be Native & Scalable, peer Association Based per Stakeholder based prior to access to the network layer (Sockets, OTT do not meet these requirements)

Addressing Scalable protocol addressing and Location, Network Address and ID separation

Transmission Significantly Reduce Packet and Access Latency

Should be Dynamically Configurable according to Context (Access Technology, Network, User Profile/ QoS, …)

Add Controls for Congestion Avoidance rather than congestion management

Ability to Enable/ Disable all Transmission Features and adjustable by Profile & Context

100% Routing &/ or Context based routing, no tunnels

Mobility Native & Scalable, with Dynamic Scope & Resolution with a Right‐sized set of Mobility Fields aim to minimise routing updates, and latency, enable lossless handovers when required (e.g. different mobility level for HST, Cars, Pedestrians and turned off for Static things)

Context Aware Built‐in, to enable intelligence to be added and to drive scalable mobility and transmission


Why change todays GTP/IPv6 access approach


Massive IPv6 header unnecessary for most sessions

No inherent user network Security (attach or association)

GTP overheads inefficient and tunnelled not routed

TCP/QUIC Transmission is inflexible and designed for fixed networks so that:

• Latency is poor for todays internet • and unlikely to be able to support AR/VR

demands over cellular

IssueInefficient use of Air interface, Inefficient use of transmission

Perpetuates many security issues in UP, includes costly IPsec tunnels over RAN-CN I/F

Inefficient processing, now one of largest costs of cellular Infrastructure

Performance limited with current algorithms and likely to remain for 5G, … difficult to tune to work well with access technologies

Cost


Example view of an NGP Architecture Evolution for INTERNET ACCESS.


Mobile

LaptopNGP IPv6IPv6

InternetNetworkPoP

Dynamic Cluster ID Mapping

Dynamic Inter‐Cluster & NGP/IP ID Mappings

NGP Cluster Router (CR) x M

Key: PPE – NGP: Packet Processing Entity == IP router plus intelligent scoping based on context

Regional intranet

Regional Regional PoP

PacketGateway(PGW)

NGP Access Router (AR) x N

Network Service Location Directory Tracks UEs and

Cluster Updates


Example NGP UP Protocol Architecture for INTERNET ACCESS


App(user)

NGP

PDCP

RLC

MAC

PHY

Bridge(PDCP<‐>Ethernet)

GatewayTCP(User)

IP(Nwk)

Eth/NwkPDCP

RLC

MAC

PHY

Eth/Nwk Eth/Nwk Eth/Nwk

LTE-Uu

RRC(DRB)

HTTP(user)

Eth/Nwk

NGP Access Router

App(user)

HTTP(user)

NGP Cluster Router

IP(User)

TCP(User)NGP

PGWBreakout GatewayF/W

Native NGP User Packet Routing


Conclusions


NGP has defined requirements for better MOBILE INTERNET ACCESS, key points:Efficiency Smaller Headers, Scalable protocol structuresSecurity Native, Scalable, Association and Stakeholder based Addressing Location, Network Address separation and ID separationTransmission Flexible, Efficient, Context‐Aware and Profile basedMobility Native, Scalable, Context‐Aware

NGP is actively encouraging SDOs to support standardisation of new protocols to improve MOBILE INTERNET ACCESS

If NGP does not succeed, many of the benefits of 5G radio access technology will be wasted in the supporting access network

The next 2‐3 years of 5G network definition time, are the LAST CHANCE TO CHANGE for an access generation!


NGP Deliverables


Web-Site: http://www.etsi.org/technologies-clusters/technologies/next-generation-protocols

Standards: https://docbox.etsi.org/ISG/NGP/Open/

GS001: NGP Scenarios Draft Approved by NGP (V1.2.1), Issued (V1.1.1),

GS002: Network Autonomics Draft Approved by NGP GR003: Routing Technologies Draft Approved by NGP GS004: ID Oriented Networks Early DraftGS005: NGP Requirements Draft in Remote ConsensusGS006: Intelligence-defined Network Early DraftGS007: NGP Reference Model Stable Draft

Whitepaper: http://www.etsi.org/images/files/ETSIWhitePapers/etsi_wp17_Next_Generation_Protocols_v01.pdf

• To join, please contact: kevin.smith@vodafone or [email protected]

foster univ - etsi · pdf filequality for best physical layer rf throughput. commercial...

Documents