IPv6 deployment scenarios in mobile networks Jouni Korhonen Netnod Spring Meeting 9-11 March, 2011 Stockholm, Sweden

Transcription

1 IPv6 deployment scenarios in mobile networks Jouni Korhonen Netnod Spring Meeting 9-11 March, 2011 Stockholm, Sweden 1 Nokia Siemens Networks

2 Foreword This presentation concentrates in most parts to 3GPP packet core; both GPRS (2G/+3G radio) & Evolved Packet System (LTE radio). 3GPP system architecture release numbers are explicitly stated when different releases make difference. 3GPP architecture has supported IPv6 since Rel-99.. and the fundamental peculiarities & flaws originate from that time (*). This presentation takes a peek into some current IPv6 deployments plans and trends I have faced when discussing with operators. The emphasis is on getting IPv6 to the end user not whole operator transmission/core/interconnection/roaming migration. 3GPP IPv6 migration guideline is in 3GPP TR Nokia Siemens Networks (*) see draft-korhonen-v6ops-3gpp-eps for a lot of details

3 In the year 2005 this was proposed for the system architecture evolution... Serving Node - C Internet Subscription AAA Registers BS All IP Access Network Serving Node - U Inter-connection Service Gateway HA Operator service network BS Had no concept of PDP Context.. Bearerless radio concept.. Had no mandatory IP Mobility as part of the architecture.. Had no mandatory tunneling.. 3 Nokia Siemens Networks

4 In the year 2011 this is what we got for the system architecture evolution... 2G 3G RAN BSC RNC Gb Iu Combi SGSN Packet Core & LTE Gn GGSN Gi LTE NodeB S1-MME MME (Gn) enodeb S1-U S11 S10 Gateway SGW PGW S5 SGi PCRF Gx Although not visible here, there are multiple IP Mobility & tunneling protocols: GTPv[12], MIPv4, DSMIPv6, PMIPv6, IPsec/MOBIKE, GRE.. APNs, default and dedicated bearers, Policy Control, PDN Connections,.. And a lot of options.. also in case of IPv6! 4 Nokia Siemens Networks

5 Some jargon and fundamentals Access Point Name (APN) is a fully qualified domain name and resolves to a specific gateway in an operator network. APN identifies the network to connect via the selected gateway. PDN (Connection) Types: IPv4, IPv6 and IPv4v6 (dual-stack). A device may have multiple PDN Connections of the same or different types open in parallel. Subscription profiles understand: IPv4, IPv6, IPv4v6 and IPv4_or_IPv6. Each IP[v4]v6 PDN Connection has exactly one unique /64. SLAAC is the only supported IPv6 configuration method for the mobile device. 3GPP separates User Plane (UP) & Control Plane (CP). Extensive tunneling for User Plane: transmission and payload IP versioning are independent of each other. GTP (UDP encapsulation) is the dominant tunneling protocol. The first real User Plane IP aware node is PGW/GGSN (hmm.. PMIPv6 is an exception). IPv6 migration solutions involving mobile host terminated tunneling strictly ruled out in 3GPP.. in standards space.. reality might prove different, though. 5 Nokia Siemens Networks

6 Mobile operators positions towards IPv6 j Driver: IPv4 address shortage Always on apps, M2M, LTEdevices,.. need a lot of addresses Growth of mobile broadband discourages NAT44 deployments Simply.. out of addresses any day soon IPv6 Driver: market demand Address shortage not an issue in the foreseen future or NATing ok React when lack of IPv6 start causing loss of subscribers May offer something to selected groups (even using tunneling) deployment Mode: aggressive.. or panic and migration Mode: it happens eventually Addresses out and NATs are...wait until Dual-Stack capability melting becomes available/justified Deploy now even if the world Accept NATing IPv4 forever around you is not up to speed Deploy selectively (i.e. end users IPv6-only is ok, NAT64 is ok. don t really care about IP version) Do not bother waiting for dualstack capable handsets and core 6 Nokia Siemens Networks

7 Introducing IPv6 in phases IPv6 rollout can be and is recommended to be phased: 1 st IPv6 at the application and end-user layer: End-user visible User Plane has technically been switch on for some time. Operators have done quite a bit of testing behind the curtains. However, Commercial offering usually involves a lot more: security(!), subscriber management/ processes, customer support, roaming, DPI, provisioning, billing, DNS, AAA, address planning,..., and a whole lot more in-house systems & databases. 2 nd IPv6 at the transport layer: Since User Plane is always tunneled there seems be no hurry to upgrade transmission to IPv6.. and sometimes transmission is owned/operator by some one else who s IPv6 rollout schedule is different. Upgrading RAN and transmission to IPv6 typically not business critical. Typically IPv6 is not the driver to touch existing IPv4 MPLS core & aggregation network. Roaming and interconnection not even defined for IPv6 yet.. it tends to work though. 3 rd IPv6 at all other interfaces: 3GPP signaling protocol information elements are IPv6 aware, even when run over IPv4. Core nodes run just fine in all IPv4 environment.. Management systems running in IPv6 typically not business critical. 7 Nokia Siemens Networks

8 Typical IPv6 network introduction plan Mobile Broadband Access as #2 Packet Core IPv4 User IPv6, (S)Gi Dual-Stack IPv6 either native or tunneled to Internet Current Mobile /64 Broadband link model makes prefix delegation for home gateways problematic (prior Rel-10) Fixed Broadband as #1 (if available) IPv6 provided to end-user Either native or tunneled 6rd (getting popular..), etc. After transition to IPv6 IPv4 can be provided with DS-Lite, some A+P Fixed Broadband Core Management Network IP Multimedia Subsystem Can be IPv6 only Backend interfaces can be IPV4 Core Network Dual-Stack Native, or Network Services obviously.. MPLS - 6PE/6VPE DNS has to be upgraded to support IPv6 access Reverse DNS Internet has to connectivity be set up Dual-Stack Network Management as last No need to upgrade at this point *if* it understands IPv6 information elements.. IMS Site Either native or tunneled Internet 8 Nokia Siemens Networks Primary colours: Supporting colours:

9 Operators and IPv6 migration; common approaches for the mobile side of operator Native Dual-Stack everywhere (3G/EPS) + NAT44: Slow paced introduction to network. Start with easy to control devices like USB dongles (and having a laptop does help migration.. dialers). Some in a mode of waiting till LTE.. but this attitude is changing. No desire to introduce NAT64 (no improvement over NAT44). No desire to introduce parallel IPv4 and IPv6 bearers.. doubles costs. Handovers between 3G and LTE are important. If pre-rel-8 SGSNs are still in use, it means IPv4 or IPv6 only bearers. IP[v4]v6 for specific use/service.. easily doable e.g. a using dedicated IPv6 APN subscription bundled with a subsidized IPv6 capable handset/dongle. 9 Nokia Siemens Networks

10 Operators and IPv6 migration; common approaches for the mobile side of operator IPv6 only + NAT64: Not overly popular but still to be reckoned with. Good for always on (LTE) handsets that are unlikely be used for active internet usage and produce huge traffic volumes.. No desire to introduce parallel IPv4 and IPv6 bearers.. doubles costs. Active push to get important applications to IPv6 native. Handovers between 3G and LTE are important. If pre-rel-8 SGSNs are used, it means IPv4 or IPv6 only bearers. Not issue with IPv6- only+nat64 approach. IP[v4]v6 for specific use/service.. easily doable e.g. a using dedicated IPv6 APN subscription bundled with a subsidized IPv6 capable handset/dongle. 10 Nokia Siemens Networks

11 NAT44 and NAT64 considerations.. NAT44 has become an important topic; something needed now and most likely forever after dual-stack deployment. Operators have accepted this.. Common to bypass NATs/FWs for heavy users and smart phones. NAT64 seen as bad as NAT44 but not equally important. 16million RFC1918 address limit has caused headache: Huge APNs where e.g. subscriber identification based on source IP address (usually some Gi box or content platform issue). Network segmentation/overlapping private networks would help but... Authorities require tracking of users behind a NAT -> real time tracing and NAT logging is becoming a real issue. 11 Nokia Siemens Networks

12 NAT[46]4 deployments.. may all exist in one network and differentiated by subscriptions IPv6-only for e.g. M2M and simple handset UE IPv6-only PGW/NAT64 NAT64 IPv6 IPv4 Internet (S)Gi Domain Dual-stack.. normal users UE dual-stack PGW/NAT44 NAT44 IPv6 IPv6 IPv4 Internet (S)Gi Domain The decision of NAT placement depends on e.g. Gateway capacity, overlapping addressing needs, Need of Gi boxes, PCC integration,.. 12 Nokia Siemens Networks

13 Dual-stack deployments.. may co-exists with NAT [46]4 and differentiated by subscriptions Dual-stack.. power users and always on smart phones UE dual-stack PGW IPv6 IPv4 Internet (S)Gi Domain The UE can also be a CPE with a cellular uplink Rel-10 introduces DHCPv6-PD.. or did someone say NAT66 or ND- Proxy? ;) CPE can do NAT44 for its internal network (remember, one IPv4 address per PDN Connection). Small business or widespot area internet solution. 13 Nokia Siemens Networks

14 Fallback scenarios and roaming cause confusion If inter-rat handovers are desired (e.g. 3G-LTE) then network migration has to be planned and subscriptions provisioned based on the lowest common nominator: Example: 3G has IPv4 only, then LTE can only support IPv4. A reason for many to wait until 3G (Rel-9 feature) and EPS (Rel-8 feature) both are IPv4v6 capable.. A dual-stack capable handset (since Rel-8) is always supposed to first try establishing IPv4v6 connection, then fall back to something different based on 1) subscription and 2) MME/PGW/GGSN configuration: See draft-korhonen-v6ops-3gpp-eps Section 8.7 for a full list of choices. There is no roaming defined yet for IP[v4]v6: GSMA has recently started working on IPv6 roaming but in general operators have not invested that much effort on it yet. It just happens to work ~75% of cases for IPv6. Real issues with inter-operator billing, thus IPv6 roaming barring is in radar. 14 Nokia Siemens Networks

15 Thank you! Questions? 15 Nokia Siemens Networks