Packet optical integration plans for OTN and MPLS switching technologies

Transcript

1. Packet optical integration plans for OTNand MPLS switching technologiesNetwork & IT Platform Strategy and Architecture© 2012 Colt Telecom Group Limited. All rights reserved.
2. Contents 1 Three separate networks today 2 Benefits to integrate L1 & L2 & L3 onto one network 3 Review of integration plans and packet optical 4 Circuits and OTN as the serving layer 5 Router by-pass and OTN versus MPLS switching2
3. A network of depth and breadth • High capacity long distance network - 35,000km • Connecting 21 countries, 39 metro networks and >100 cities • 19 (+1) data centres and 18,000 connected buildings3
4. Network layer integration – Historical situation L2 CPE L2 service L2 PE L2 packet core L2 access & aggregation IP NNI L2 serviceL3 service L2 CPE L3 CPE L3 packet core L3 PE NB: Colt managed CPE (not customer’s) IP services over Ethernet access (metro) 4 Ethernet services (metro & inter-metro)
5. Benefits associated with network integration Simplification Operations, architecture, service nodes (PEs) Reduced CAPEX (less devices per service, “pay as you grow” Cost core) and OPEX (simplified delivery & assurance) Technical Statistical multiplexing gain in the core Improved service unit costs (CAPEX/OPEX reduction, Product statistical multiplexing), better delivery lead time and TTR, L2 & L3 service blending (Integrated Routing & Bridging on PE)5
6. Network layer integration – L2/L3 CPE L2 CPE L2 service L2 PE L2 packet core L2 access & aggregation IP NNI L2 service L3 service L2 CPE L3 packet core L3 PE L3 features moving to the PE layer6
7. Network layer integration – L2/L3 edge L2 packet core L2 CPE L2 service L2/L3 PE L2 access & aggregation L2 service L3 packet core L3 service L2 CPE7
8. Network layer integration – L2/L3 packet optical core L3 features moving to the L3 features L2 CPE core layer / cloud L2 service L2/L3 PE L2 access & L2/L3 packet optical core aggregation L2 service L3 service L2 CPE8
9. Vision on the best packet optical platform • Multi-layer switching platform – WDM (colour-less, contention-less, etc.) – OTN – Packet • OTN switching L2/L3 packet optical core – Fill-in the high speed waves • Packet switching – MPLS switching (LSR) – No LER (VPN, VPLS, GRE, MC), no BGP – CP protocols (IS-IS, OSPF, TE, LDP, RSVP)9
10. Network layer integration – Design thoughts Network design considerations •L2 and L3 services integrated on the same physical infrastructure but kept logically separated through the use of dedicated VLANs – Service QoS characteristics uncompromised thanks to no queue sharing (8xL2 queues + 8xL3 queues) •Key protocols for L2 and L3 services Functions L2 services L3 services IGP OSPF IS-IS Path computation NMS/OSS based IGP (core) Topology discovery OSPF-TE (booked bandwidth) IGP Path creation ERO + RSVP-TE IGP + LDP •No mandatory requirement of MPLS-TP for L2 services – To start with MPLS properly tuned seen good for the job (OAM included) – … but over time MPLS-TP might be useful for service assurance activities10
11. The reality of circuits versus packets at Colt • Demand for SDH services (2.5 and 10Gbps) still exists – No foreseeable time for end of sale (not within the next 2 years) – But largely outnumbered by Ethernet services (1 and 10Gbps) • No demand for OTN services (wholesales) exists – And no early signs this is going to change – But OTN switching capability (as an internal network feature) remains attractive • High speed Ethernet services (1 and 10Gbps point-to-point) are provisioned over the optical transport layer – But 1 and sub-10Gbps are being moved to the MPLS packet transport layer as scale of this layer increases • Sub-1Gbps Ethernet, IP and VoIP services are provisioned over MPLS – Integrated packet optical core to become more prominent in the future Proportion of circuits decreasing over time11
12. Circuits supported by different flavours of OTN • OTN framing – Proven track record with G.709 digital wrapper with FEC – Even used in packet platforms (IP over WDM with coloured optics) • OTN multiplexing – Must have LO ODU to efficiently fill-in waves in point to point topology 1G circuits 10G wave12
13. Circuits supported by different flavours of OTN • OTN framing – Proven track record with G.709 digital wrapper with FEC – Even used in packet platforms (IP over WDM with coloured optics) • OTN multiplexing – Must have LO ODU to efficiently fill-in waves in point to point topology • OTN switching – Useful to more efficiently use waves in transit situations (bus topology)13
14. Circuits supported by different flavours of OTN • OTN framing – Proven track record with G.709 digital wrapper with FEC – Even used in packet platforms (IP over WDM with coloured optics) • OTN multiplexing – Must have LO ODU to efficiently fill-in waves in point to point topology • OTN switching – Useful to more efficiently use waves in transit situations (bus topology)14
15. Router bypass and circuit versus MPLS switching What type of inter-PE connectivity on transit sites? OTN P PE service tunnel MPLS transport tunnel service VLAN15
16. Router bypass and circuit versus MPLS switching What type of inter-PE connectivity on transit sites? •No router bypass but MPLS switching on transit nodes – Simplified architecture hence simplified operations (less tunnels, one case policy) – Only marginal extra costs as co-located OTN and LSR functions – Statistical multiplexing benefit (assuming minimum traffic variability) – Additional latency reasonably negligible – Potential scale concerns for large tier-1 •Router bypass i.e. circuit switching on transit nodes – Challenge to map inner LSP transport tunnel to separate circuit (burn dedicated ports on P and OTN as an expensive mitigation step) – Circuit bandwidth tax (more controllable if using low granularity ODUFlex) – Better latency in general (unless store & forward mapping scheme used) Router bypass not seen too hot for Colt … if technically achievable16
17. Summary • Packet transport layer likely to become the pivotal transport layer – Services run packets and use only 1/15th of their contracted bandwidth • Great stories in optics to bring in 100G+, SD-FEC, flexible grid, etc. – Unclear rationale to massively scale the optical and OTN layers if the network would run no actual traffic • Different implementation scenarios for packet optical – LSR switching on optical core (with packet CP) or, in a later timescale, Software Defined Networks controller (ONF or IETF definition) MPLS switching: yes OTN switching (infra): yes today Decision factors: demand, scale, pricing17
18. Thank you. [email protected]© 2012 Colt Telecom Group Limited. All rights reserved.