IPv6 in the Telco Cloud - APNIC · As IPv4 addresses deplete, it will be more expensive to extend IPv4 resources Dual-Stack is an effective transition technology but does not solve

Telstra Unrestricted Copyright Telstra©

IPv6 in the Telco Cloud

APNIC 44 – September 2017

Sunny YeungSenior Technology Specialist, Telstra Wireless Network [email protected]

Technical lead forWireless IPv6 deploymentWireless Mobile Edge and Telco Cloud IP Architecture

Introduction

2 | Telstra Unrestricted | IPv6 in the Telco Cloud | Sunny Yeung | 09/2017

What does a IPv6 Mobile Network look like?

Where are we at with deployment?

What challenges lies ahead? Where else do we need to consider IPv6?

Executive Summary



IPv6 for a Mobile Network


As IPv4 addresses deplete, it will be more expensive to extend IPv4 resources

Dual-Stack is an effective transition technology but does not solve the IPv4 depletion problem

Introducing IPv6:• Reduced dependency on NAT• Remove the need for regionalisation• Pushes applications to move to IPv6

IPv6 for a Mobile NetworkWhy?


IPv6 for a Mobile NetworkTraffic Flow


IPv4PublicInternet

CarrierNetwork(IPv4+IPv6)

IPv6PublicInternetIBR

NAT64PublicIPv6toPublicIPv4

NativeIPv6

RadioNetwork(IPv4transport)

EPG

Single–BearerIPv6onlyuserplane

IPv6

Two existing APNs – one for Handsets, one for Mobile Broadband and Tethering

or

464XLAT + NAT64 + DNS64 for the Handset APN only

IPv6 enabled DNS for all other APNs

IPv6 for a Mobile NetworkDifferent APNs for different purposes


Telstra.WAP

Telstra.Internet

NAT64/DNS64

464XLAT Internet

DNS-DS/NAT44 Internet


What have we achieved so far?


Most major content providers are providing IPv6 reachability

Some providers have mandated IPv6 support in their app stores and all new apps must support native IPv6

Networks are evolving to support new IPv6 connections – SS or DS

Devices are beginning to support IPv6 natively





Source - https://www.google.com/intl/en/ipv6/statistics.html

So what is there left to do?



Complexity in the network


Value Added Services or packet processing requires manual steps today

Each path may require a unique VPN or PBR to define its path. The more services, the more convoluted this becomes.

Complexity in the network


InternetMobileNetwork

PGW

FW VideoOptimizer

ParentalControl

BorderRouter

Complexity in the networkNetwork Function Virtualization


Virtualisation may reduce the cost of such network functions, but does not necessarily reduce the network complexity

Each path may require a unique VPN or PBR to define its path. The more services, the more convoluted this becomes.

vFW vVideoOptimizer

vParentalControlvPGW

InternetProviderMobileNetwork

PE

Leaf

BR

vCGN

Leaf

Spine Spine

Complexity in the networkSDN only solves one of many problems


Software-Defined Network means many different things to many people.

Service Function Chaining is the equivalent of PBR for SDN.

1. Identify the user2. Allocate a service chain to the

user according to their subscription

3. Program the network to direct the packets of this user to the appropriate VNFsvFW vVideo

OptimizervParentalControlvPGW


PE

Leaf

BR

vCGN

Leaf

Spine Spine

IfServiceAthenvFWthenvCGNthenInternet

IfServiceBthenvFWthenvParentalControlthenvCGNthenInternet

IfServiceCthenvVideoOptimizerthenvParentalControlthenvCGNthenInternet

OpenflowSwitch

Let’s dive deeper…


Complexity in the networkUnderlay and Overlay


Underlay can be IPv4 only. But why?

Can we do Point-to-Point /127 and do away with IPv4 altogether?

Is VXLAN over EVPN the best solution for a Telco Cloud? What about service chaining?

This should all be contained within a single AS. What routing protocol is run between VTEPs?

If this is entirely IPv6, can we leverage Segment Routing for SDN control?

vFW vVideoOptimizer

vParentalControlvPGW


PE

Leaf

BR

vCGN

Leaf

Spine Spine

SR?

Complexity in the networkUnderlay and Overlay


Openstack is somewhat ready for IPv6 – neutron supports IPv6 subnets and DHCPv6, as well as SLAAC

You can begin to onboard tenants that can be IPv6 enabled.

Really there is only two options –Dual-Stack everything, or Conditional Dual-Stack.

API endpoints

Databases

Automation

Interfaces

IPv4/IPv6

IPv4/IPv6

IPv4/IPv6

IPv4/IPv6

Does not support the use of an external IPAM module for IPv6 to allocate the subnet.

OVS tunnel types do not support IPv6 endpoints until OVS 2.6.0+

Single-stack IPv6 tenant networking not fully available.

VPN support not good until at least Kilo.

NAT on OvS for IPv4.

Complexity in the networkOpenstack does not fully support IPv6



Fog and Transport


Fog


5G requires super low latency depending on the network slice required and the application behind it.

At some point, we must move the cloud even closer to the user

Telco Cloud:1. Service Functions in VNFs2. Virtualised Packet Core Nodes3. Media Cache4. Value Added Services

TransitProvider ContentProvider

MPLSCore

Internet

Aggregation

PE

BorderRouter

RoamingGateway

CRAN TelcoCloud

Transport


IPv6 SS for transport layer ? (6PE)

SRIPv6 to replace traditional MPLS?

6VPE required to ensure services provided by different VPNs have IPv6 support

TransitProvider ContentProvider

MPLSCore

Internet

Aggregation

PE

BorderRouter

RoamingGateway

CRAN TelcoCloud


What is our goal?


The full removal of IPv4 private and public addressing from the entire network

Enable NAT translations for only what is necessary

IPv4aaS

What is our goal?


There is a long journey yet for IPv6, this is just the start…



Thanks!


IPv6 in the Telco Cloud - APNIC · As IPv4 addresses deplete, it will be more expensive to extend IPv4 resources Dual-Stack is an effective transition technology but does not solve

Documents