LACNOC Technology Options for Access Providers with IPv6. Michael De Leo CTO LATAM

Transcription

1 LACNOC 2010 Technology Options for Access Providers with IPv6 Michael De Leo CTO LATAM 1

2 Agenda Service Provider Access IPv6 in Access Networks xdsl, ETTH, WLAN and CABLE So where is the content? IPv4 Exhaustion Technologies Applying Technologies to Migration Paths Your Options Conclusions 2

3 Service Provider: Access 3

4 IPv6 Broadband Access Solutions Layer 2 Encapsulation(s) Dial PSTN NAS IPv4/IPv6 Firewall PIX, Cisco IOS FW ISP A Internet DSL DSLAM BAS Enterprise Access Ethernet Mobile Cable DOCSIS 3.0 Proposal RAN Head-End IPv6 Prefix Pools IPv6 RADIUS (Cisco VSA and RFC 3162) DHCPv6 Prefix Delegation Stateless DHCPv6 DHCPv6 Relay Generic Prefix Distributed Computing (GRID) Video IPv6 Multicast ATM RFC 1483 Routed or Bridged (RBE) PPP, PPPoA, PPPoE, Tunnel (Cable) Dual-Stack or MPLS (6PE) Core IPv4/IPv6 4

5 Two Broadband Access Models Today Network access provider = internet service provider Access Layer Core ISP Network access provider # internet service provider Access Layer Core NAP ISP 5

6 Transition Technologies: Summary Today Private IP/ NAT IPv6 over IPv4 (6rd/ 6PE) Dual-Stack IPv4 over IPv6 (DS- Lite) All IPv6 Business / Consumer IP NGN Prosper Prepare Preserve = IPv4 = Private IP = IPv6 6

7 xdsl, ETTH and WLAN Networks 7

8 Point-to-Point Model Customer Premises NAP NSP DSLAM (NAP) Layer 2 Edge Router xdsl Customer Premises NSP Acc Sw Agg Sw (NAP) Layer 2 Edge Router ETTH 8

9 L2TPv2 Access Aggregation (LAA) Model Customer Premises NAP NSP DSLAM PPP BRAS NAP L2TPv2 Edge Router ISP xdsl AAA RADIUS Server Acc Sw Agg Sw BRAS NAP Edge Router ISP PPP L2TPv2 ETTH AAA RADIUS Server 9

10 PPP Terminated Aggregation (PTA) Model Customer Premises NAP NSP DSLAM BRAS Edge Router PPPoX NAP ISP xdsl AAA RADIUS Server Acc Sw Agg Sw PPPoE BRAS NAP Edge Router ISP ETTH Only one PPPoA session per PVC Multiple PPPoE sessions per VLAN The PPPoE sessions can be initiated by the hosts or the CPE AAA RADIUS Server 10

11 Hybrid: IPv4 LAA Model and IPv6 PTA Model Customer Premises NAP NSP DSLAM BRAS NAP Edge Router PPP PPP PPP L2TPv2 ISP xdsl AAA RADIUS Server Acc Sw PPP Agg Sw PPP PPP BRAS NAP L2TPv2 Edge Router ISP ETTH AAA RADIUS Server IPv4 IPv6 11

12 IPv6 RBE Different Than IPv4 RBE: Pick out the 0x86DD type and route the traffic Enabled per PVC, IPv6 address is configured per PVC, each PVC supports a different subnet IPv4 Traffic Bridged BRAS L2TPv2 ISP IPv6 RBE IPv6 Traffic Edge Router IPv4 IPv6 12

13 Cable Networks 13

14 CableLabs IPv6 Decision and Approach CableLabs members put IPv6 in consideration for DOCSIS 3.0 Cisco responded with proposal for IPv6 architecture and features IPv6 identified as one of top three ranked order priorities by MSOs Decision: DOCSIS 3.x MUST fully support IPv6 Cisco primary author for DOCSIS 3.0 IPv6 and enhanced IPv4/6 Multicast specifications Rationale Increased address space for CM management New CPE services Proposed phases Phase 1 CM hardware impacting features, CM provisioning and management over IPv6, embedded IPv6 router in CM Phase 2 remaining IPv6 features for CPE services, for example IPv6 CPE provisioning and IPv6 service support 14

15 IPv6 Deployment Models for DOCSIS 3.0 Customer Admin Domain Access Model 1 MSO Admin Domain Servers DHCP, DNS CPE1 Access Model 2 CPE2 HOME/ SMB Access Model 3 CPE Router CM1 Bridge CM2 Bridge HFC CMTS Router CORE TFTP TOD Management To Internet CPE3 HOME/ SMB CM Router Management Prefix: 2001:DB8:FFFF:0::/64 Service Prefix: 2001:DB8:FFFE:0::/64 Customer 2 Prefix: 2001:DB8:2::/48 Customer 3 Prefix: 2001:DB8:3::/48 HFC Link; Assigned 2001:DB8:FFFF:0::/64 (Mgmt) and 2001:DB8:FFFE:0::/64 (Serv) Customer 2 Premises Link; Assigned 2001:DB8:2:0::/64 Customer 3 Premises Link; Assigned 2001:DB8:3:0::/64 Routers Span Customer and MSO Administrative Domains 15

16 So where is the content? 16

17 Content on IPv4 today Not much content in the IPv6 world yet Customer equipment mostly IPv4 devices, but more devices are IPv6 ready Not all client CPE or hosts have same basic functions or migration technologies Most users have windows computers, but more than just computers connect to the Internet Tivo, Xbox/PS, When IPv4 address are exhausted what are some addressing scenarions: IPv4 NAT (sometimes called NAT44) IPv6 native IPv6 with Address Family Translation (AFT) 17

18 How to get to content: IPv4 IPv6 We will need NATting or more precisely Address Family Translation (AFT) NAT-PT (NAT Protocol Translation) is deprecated New proposals are being developed under the Behave and Softwires Working Groups in the IETF See draft-wing-nat-pt-replacement-comparison-02 Still in flux many proposals are being dropped Different places where AFT can take place CPE router ISP Edge Network Core 18

19 Challenges with NAT and AFT NAT or AFT has its own challenges Scalability e.g. applications like Google Maps (AJAX) use many ports simultaneously May require application changes to work with NAT/AFT Stateful management could be a challenge lawful intercept and logging RFC 2993 documents some architectural implications of NAT Some cases might require DNS rewriting 19

20 Network Address Translation (NAT) Terminology NAT44 NAT444 NAT46 NAT464 NAT64 NAT66 IVI The classic IPv4 NAT Double NAT (NAT on Residential Gateway (RG) + NAT within SP network) Protocol translation from IPv4 to IPv6 (may also include DNS46) Double NAT with IPv6 transport Protocol translation from IPv6 to IPv4 (may also include DNS64) Hiding addresses for reachability or domain independence Prefix-specific & stateless address mapping for IPv4/IPv6 coexistence and transition LSN Large Scale NAT In practice, all three mean the SP CGN Carrier Grade NAT performs some form of NAT for many subscribers AFT Address Family Translator 20

21 Should You Do Large Scale NAT (LSN) or IPv6? This is not an either/or choice. Depending on your drivers, you may have to do both. Long term, not doing IPv6 increases your risk/cost to support any new services, or changes to existing services For now you may also choose do neither if you have plenty of IPv4 addresses But in the long run there is risk. Competitors scrambling today might get the jump using IPv6. Still, not being forced unwillingly to the market first is a nice position :-) 21

22 Large Scale NAT444 in Operation Subscribers Access Provider Network Public Internet IPv4 Today: IPv4 IPv4 public Public IPv4 Internet IPv4 Core IPv4 Adding LSN: IPv4 (NOT)-IPv4 public LSN Public IPv4 Internet IPv4 Core = public IPv4 = NOT public IPv4 22

23 LSN 444 Advantages and Challenges Advantages Addresses immediate IPv4 exhaust problem No change to subscriber CPE No IPv4 re-addressing in home Allows deferral of IPv6 architecture selection Dense utilization of Public IP address/port combinations No new UNI protocols extensions No standard body delays Challenges User traceability Local traffic requires hair pinning Telling users you are out of bindings Users expectations versus having a private IP service only? Margin & competitive implications Operations & troubleshooting of transient issues User control over NAT 23

24 Large Scale NAT 444 Will Happen Its simply too late for IPv6 to meet IPv4 exhaust demands gracefully IPv6 Content Tilt Public IPv4 BB users Private IPv4 BB users Public IPv6 BB users Accessing IPv6 Internet * source: Infonetics

25 LSN 444: Who Can, Who Can t Who First? What Costs? Simplistic answer: where the growth of public endpoints outstrips capacity Mobile (Emerging) Mobile (Developed) Wireline Addressable Endpoints (Developed) Incremental costs of Network based NAT Capital Operations OSS Binding maintenance for application keep alives Lawful Intercept binding archiving But what applications cannot afford LSN? Applications requiring universal connectivity / addressability at low cost Applications of high bandwidth & transport little revenue ($Cost / Mb) 25

26 LSN Must Handle Traffic Growth Mobile Data Traffic In 2010 Internet video will surpass P2P in volume. This will be the first time since 2000 that any application has displaced P2P as the top traffic driver. Almost 64 percent of the world s mobile traffic will be video by Mobile video will grow at a CAGR of 150 percent between 2008 and Mobile video has the highest growth rate of any application category measured within the Cisco VNI Forecast at this time. Mobile broadband handsets with higher than 3G speeds and laptop aircards will drive over 80 percent of global mobile traffic by A single high-end phone like the iphone/blackberry generates more data traffic than 30 basic-feature cell phones. A laptop aircard generates more data traffic than 450 basic-feature cell phones. Source: 2009 Cisco Global Mobile Data Traffic Forecast Update Subscriber Traffic Growth Must Be Matched Against Expected Trends in Service Delivery Price/Performance 26

27 LSN Engineering Depends on Service Mix Price per Megabyte for different services What will be your LSN engineering constraint: Bandwidth? Bindings? What is your margin for the service most constrained? Your answer is one indicator of when IPv6 becomes a must 27

28 Choosing an Optimal Address-Type per Service Service Characteristics wrt/ Addressing Type IPv4 IPv4 over LSN IPv6 Cisco s Global Consumer Internet Traffic Forecast HTTP/ Most content is currently IPv4 Client- Server Walled Garden IPTV Cost/bit optimized delivery Client- Server Over-the-top HD Video Cost/bit optimized delivery Client- Server BitTorrent Availability of seeds without keep-alive Client- Client Sensors & Alarms Low power/utilization devices can t afford frequent keep-alives Client- Server Telepresense & Video Conferencing Specialized P2P ; Multiple applications linked together; strong requirements for ALGs if NAT is employed Client- Client Internet-to-TV traffic will bypass Internet video-to-pc traffic by 2009 and will exceed 1 exabyte per month in Approaching the zetabyte era, Cisco 2008 * Assessment based on the assumption that LSN will incurr additional cost (CAPEX for additional NAT-device and/or processing, as well as OPEX for NAT operation & control) 28

29 Access Network Provider Services Model Affects IPv6 Approach Integrated Services Provider Seek to create user experience and suite of own services Often independent networks for different services Cost per bit is not 1st concern IPv6 mainly for internal optimization in the short term. Connectivity Services Provider Optimize cost of bandwidth delivery and scale. One network for all services Support with multiple applications that drive connectivity IPv6 as a differentiator 29

30 An LSN + IPv6 SP Adoption Scenario One of the Many Possible... IPv6 enabled endpoint IPv4 enabled endpoint IPv4 address pool exhausted New end-systems deployment (handset/rg) IPv6 only endpoints technically viable IPv6 Internet (IPv6 only transport viable from a market perspective) NAT64 Dual-Stack deployment NAT46 Large Scale NAT (LSN) introduction time 30

31 IPv4 Exhaustion Technologies 31

32 Menu of IPv4 Exhaust Technologies Method 1 Method 2 Method 3 Method 4 Method 5 Method 6 v6 Hosts (& Dual Stack) Large Scale NAT 444 NAT 64 v6 Tunneling v4 Tunneling over v6 v4 Subnet Trading / Exchange Interworking / coexistence will be necessary 32

33 Method 1 IPv6 Hosts (and Dual Stack) v6 hosts: Relief of tight private IP address space Bypass NAT application problems Functional separation of firewall from universal addressability v6 only devices likely to start in application specific pockets (sensors, SIP phones) v6+v4 hosts (dual stack): Will smooth market entry as devices & network are less tightly coupled Defers some IPv6 infrastructure investment (e.g., use DNSv4) Operation costs increase for Dual-Stack We have forgotten how to operate multi-protocol networks Facts: Obvious: There aren t enough IPv4 addresses to sustain the current v4 model Corollary: There aren t enough IPv4 addresses to support the dual stack model 33

34 Method 2 Large Scale NAT 444 NAT is already helping with v4 exhaustion Works via mapping multiple hosts into the port range of a single IP-Address Creates client reachability & potential ALG issues (sometimes sold as a feature ) Expanding NAT into SP infrastructure increases longevity of IPv4 Sadly client software has little reason to minimize number of Ports used Large scale carrier NAT not considered in application design Ultimately this will be a scaling limitation Cisco s FWSM & ACE Enterprise NAT boxes already doing LSN in Mobile Smartphone applications often different than those on Broadband PCs Changing usage patterns for mobile (wireless as the only connectivity) All NAT solutions suffer from complicating host-host communications. 34

35 Method 3 NAT 64 IPv6 hosts to access any IPv4 Internet endpoint Some IPv6 devices might only need sporadic access to the IPv4 Internet Solution: Global NAT 64 with N:1 mapping (i.e., Large Scale NAT 64) IPv6 hosts to Private IPv4 servers Local domain might allow v6 clients to access content on local v4 hosts Access to IPv4 home server while on the IPv6 road Solution: Local NAT 64 with 1:1 mapping Application Layer Gateways (ALG): Application specific NAT Session Border Controller ALG (for SIP phones) DNS ALG Fact: Going from 6 4 is easy. Initiating from the IPv4 side is harder. 35

36 Method 4 v6 Tunneling : v6 over PPP SPs would love to have their embedded access infrastructure support IPv6 However legacy DSLAMs often cannot pass IPv6 These DSLAMs can pass PPP or IPv4, so it is possible to tunnel IPv6. This means massive investment reused Tunnels can originate from RG or CPE. When on CPE, no coordination with RG or Access Provider required! 36

37 Method 4 v6 Tunneling: IPv6 Rapid Deployment (6rd) A form of v6/v4 which efficiently traverses the aggregation cloud without added IPv6 provisioning For IPv6 traffic destined for the Home, the 6rd Relay pulls the RG s IPv4 from within the destination IPv6 address (when combined with a CGN lookup this is 6rd+) For IPv6 traffic destined to a nearby 6rd user, the RG pulls the target IPv4 tunnel endpoint from within the destination IPv6 address 6rd Relay 6rd RG RG IPv4 Address Residence s IPv6 Subnet is constructed from: ISP s IPv6 Prefix + RG IPv4 Address + Interface ID For IPv6 traffic destined to the backbone, the RG uses the destination IPv4 of the 6rd Relay. Backbone traffic is identified by masking the destination IPv6 for either: NOT (ISP IPv6 Prefix), or NOT (routable unicast public IPv4 address) /32 /64 /128 37

38 Approach: Keep the IPv4 Access & Aggregation Network as is AAA, DHCP, OSS IPv6 Ready Backbone (6PE or Native) IPv6 Ready Hosts RG Access Node (DSLAM) BNG (BRAS, CMTS) Router IPv4-Only Access, Aggregation, AAA 38

39 Tunneling 6rd RFC5969 IPv4 + IPv6 IPv4 + IPv6 IPv4 + IPv6 Subscriber IPv6 prefix derived from IPv4 address 6rd One line global config for IPv6 Gateway 6rd Dual Stack Native or 6PE Core 6rd CE 6rd Border Relays 6rd IPv4 Native dual-stack IP service to the subscriber Simple, stateless, automatic IPv6-in-IPv4 encapsulation & decapsulation IPv6 traffic automatically follows IPv4 routing 6rd Border Relay placed at IPv6 edge Conceptually similar to 6to4 (RFC3056) 39

40 6rd Automatic Prefix Delegation (From a Global IPv4 Prefix) 6rd IPv6 Prefix Customer IPv6 Prefix 2011: Subnet-ID Interface ID This prefix length is variable in 6rd, /28 is just an example Customer s IPv4 prefix (32 bits - or less) 40

41 6rd Automatic Prefix Delegation (From a Private IPv4 Prefix) 6rd IPv6 Prefix Customer IPv6 Prefix 2011: Subnet-ID Interface ID Customer s IPv4 prefix Without the "10." (24 bits) 41

42 Packet Flow and Encapsulation 6rd 6rd 6rd IPv4 + IPv6 IPv4 + IPv6 IPv4 + IPv6 RG 6rd Border Relays IPv4 + IPv6 Core 6rd IPv4 IF (6rd IPv6 prefix) THEN encap in IPv4 with embedded address Dest = Inside 6rd domain IPv6 Dest = Outside 6rd domain ELSE 2001: :0101 Interface ID Not 2001: ENCAP with BR IPv4 Anycast address Interface ID 42

43 6rd BR Setup and Provisioning 6rd BORDER RELAY REPRESENTATIVE CONFIG (IOS, ASR1K) interface Loopback0 ip address ! interface Tunnel0 NAT44 tunnel source + 6rd Loopback0 tunnel mode ipv6ip NAT 6rd IPv4-Private tunnel + 6rd IPv6 ipv4 prefix-len 8 Native tunnel Dual Stack 6rd prefix 2001:db80::/32 Access ipv6 to Customer address 2001:db80:6464:100::/128 Node BNG RG (IPv4) (IPv4) anycast! ipv6 route 2001:db80::/32 Tunnel0 ipv6 route 2001:db80:6464:100::/56 Null0 IPv4-only AAA and/or DHCP 6rd Border Relay 1. Simple BR must and Easy have to IPv6 setup! reachability (Native, 6PE, GRE Tunnel, etc). Additional tip: Use the ipv6 general-prefix command to automatically create the ipv6 address anycast 2. address... An access-network-facing IPv4 address (likely anycast) 3. Global 6rd ISP Prefix and Length *One BR may serve one or more 6rd Domains 43 IPv4 IPv6 + IPv4

44 Method 5 v4 Tunneling over v6 For networks willing/able to push IPv6 access & aggregation Use #1: SPs exhausting 10.x.x.x space in their aggregation networks Cable Modems & mobile access/aggregation devices: NB/eNB, S-GW,... Use #2: IPv4 & IPv6 services over IPv6 transport Dual Stack Light NAT 444 by SP (1:1 or N:1) means no impact to premises IPv4 numbering Allows graceful turn-down of IPv4 over time CPE 44

45 Method 6 v4 Subnet Trading / Exchange Markets form to balance offer & demand for scarce resources Discussions on IPv4 subnet trading already started (as did the trading in some cases) Current IPv4 address allocation mechanisms were not built to support the dynamic reallocation of subnets Facilitating address trading means protecting against address hijacking / false announcements etc. BGP Prefix Validation draft-pmohapat-sidr-pfx-validate Source: can-an-ipv4-stock-market-stave-off-address-depletion-ipv6.ars 45

46 Your Options 46

47 IPv6 Dialogs with Broadband/Mobile Carriers Business Driver Interest UNI Exhaust Technologies Broadband: reuse access infrastructure Broadband: replace access infrastructure Mobile: IPv6 Handsets Method Dual Stack IPv6 Hosts 1 Method 2 Method 3 Method 4 LSN 4 4 LSN 6 4 IPv6 over PPP / 6rd+ Method Dual Stack 1 IPv6 Hosts Method LSN 4 4 LSN 6 4 Dual Stack Lite: v4/v6 Method IPv6 IPv6 Hosts 2 Method 3 Method 5 1 Method 3 LSN 6 4 Note: This table doesn t cover v4 only topologies, ie: 2 LSN 4 4 or 6 v4 Subnet trading Method Method 47

48 Making the Architectural Choices The good news: You have lots of Options! (The bad news: you have a lot of options!) Aggregation Native over MPLS L3 Adjacency One BNG Many BNGs Wholesale Tunneling CPE v6 / Ethernet v6 / PPPoE RG UNI Routing Bridging Access v6 / Ethernet v6 / PPPoE v6/ v4 L2 Edge 1:1 VLAN N:1 VLAN Additional complexities: - Integrating embedded IPv4 - Transition mechanisms Which mix is right for you? 48