Layer 1, 2 and 3 Integration

1. Layer 1, 2 and 3 Integration Vision and ongoing developments

2. Contents

3. Introduction Colt 13 countries, 34 EU MANs and 24 EU/USA/CEE POPs 19 datacentres and 16000 on-net building Served by a 25000km operated fiber network Next-Generation Networks and Services Optical Transport Carrier Ethernet IP/MPLS (Internet and IPVPN) IMS/VoIP Cloud/xAAS

4. L2 & L3 product requirements L2 & L3 product requirements (table) Service type Protection QoS (#classes, hard QoS for CIR, CAC) Latency (algorithmic trading, vs. hop count, how to control it, static routing with ERO/NMS vs. IGP) L2 & L3 product requirements (table) Service type Protection QoS (#classes, hard QoS for CIR, CAC) Latency (algorithmic trading, vs. hop count, how to control it, static routing with ERO/NMS vs. IGP)

5. Historical situation NB: Colt managed CPE (not customer’s) No CPE integration => 2 box model to deliver all L3 services No PE integration => two distinct PE devices and platforms (different vendors) No P integration => two inter-metro backbone built typically with unprotected Ethernet core links over LH DWDM But L2 access & aggregation already integrated since the beginning of MSP and earlier access transports (SDH, EoS) No CPE integration => 2 box model to deliver all L3 services No PE integration => two distinct PE devices and platforms (different vendors) No P integration => two inter-metro backbone built typically with unprotected Ethernet core links over LH DWDM But L2 access & aggregation already integrated since the beginning of MSP and earlier access transports (SDH, EoS)

6. Background to the L2 & L3 separation In the mid-long term as per MSP business case 2007, given the amount of traffic planned over MSP inter-metro backbone, the only viable options foreseen are DWDM LDN wavelength at 10G for tier-1 cities and 1G unprotected LDN Ethernet for smaller tier-2 cities. Despite highly appealing to achieve network convergence, IP/MPLS core Ethernet option was not retained for several main reasons: No sub-50 ms protection (see MPLS FRR and BFD) No support to hard QoS and dual colour rates (i.e. EIR<>CIR) No point & click, and service provisioning continuity (see ASPEN) Low amount of free capacity left for LANLink services towards tier-2 IP cities Cost model and price points (see high price of GSR access port, plus requirement to double Cisco discount rate from 40% to 80%) Readiness timeframe of IP Core layer (see MSP to be introduced before early 2008)In the mid-long term as per MSP business case 2007, given the amount of traffic planned over MSP inter-metro backbone, the only viable options foreseen are DWDM LDN wavelength at 10G for tier-1 cities and 1G unprotected LDN Ethernet for smaller tier-2 cities. Despite highly appealing to achieve network convergence, IP/MPLS core Ethernet option was not retained for several main reasons: No sub-50 ms protection (see MPLS FRR and BFD) No support to hard QoS and dual colour rates (i.e. EIR<>CIR) No point & click, and service provisioning continuity (see ASPEN) Low amount of free capacity left for LANLink services towards tier-2 IP cities Cost model and price points (see high price of GSR access port, plus requirement to double Cisco discount rate from 40% to 80%) Readiness timeframe of IP Core layer (see MSP to be introduced before early 2008)

7. Technical and business benefits General simplification: operational, architecture, service nodes (PEs) Statistical multiplexing with improvement of the service unit costs Product: L2 & L3 service blending with IRB General simplification: operational, architecture, service nodes (PEs) Statistical multiplexing with improvement of the service unit costs Product: L2 & L3 service blending with IRB

10. L2 & L3 integration phases Phase 1: Access (CPE) Phase 2: Edge Phase 3: Core Alternative phasing?

11. Access (CPE) integration No CPE integration => 2 box model to deliver all L3 services No PE integration => two distinct PE devices and platforms (different vendors) No P integration => two inter-metro backbone built typically with unprotected Ethernet core links over LH DWDM But L2 access & aggregation already integrated since the beginning of MSP and earlier access transports (SDH, EoS) No CPE integration => 2 box model to deliver all L3 services No PE integration => two distinct PE devices and platforms (different vendors) No P integration => two inter-metro backbone built typically with unprotected Ethernet core links over LH DWDM But L2 access & aggregation already integrated since the beginning of MSP and earlier access transports (SDH, EoS)

12. Edge integration Access integration is not a pre-requisite to Edge integrationAccess integration is not a pre-requisite to Edge integration

13. Core integration Access integration is not a pre-requisite to core integration, but Edge isAccess integration is not a pre-requisite to core integration, but Edge is

14. L2 & L3 integration phases and challenges Access Quick win phase to save the L3 CPE at the customer premises Challenge to identify the eligible L3 product feature set The backup, resilience and remote access options would require a L3 CPE Edge Optimisation of the service node architecture PE shared for L2 & L3 services and better deployed to some selected places Challenge to meet the L2 product requirements Core Ultimate and most complex phase to happen after the edge integration Higher challenge to meet the L2 product requirements

15. Dual-vendor requirement General strategy at Colt for the main platforms like L2 and L3 Step 1: Launch with the main vendor Step 2: Close the critical gaps of the second vendor Step 3: Launch the second vendor Multiple direct benefits… Stronger influence on the vendors’ roadmap Stronger influence in the business reviews (e.g. price lists) Risk mitigation for vendors going bankrupt But other great challenges generated More complex operations for service delivery, service activation and service assurance Overlay deployment and patches due to limited inter-working (cf. proprietary and different standard implementations) Infrastructure costs (power and cooling) and service distribution

16. Dual-vendor requirement What it means for the L2/L3 integrated solution

17. The OSS & BSS environments: service delivery, service activation and service assurance Umbrella Provisioning Tool (UPT) framework

18. Progress review of the Colt project Where we are now On the company roadmap with divisional resources already engaged Colt-wide integration work to kick off mid 2010 High level design done without major development required from the main vendors Various development pieces to come from the general roadmap Areas under investigation Strategy for the historical L2 & L3 edge and core: cap and grow, natural decline over time (ceases), migrate PE architecture: shared PE for L2 & L3 (possibly multisegment-PW with reduced PE footprint) UPT launch: tactical, mass production

19. Progress review of the Colt project Where we are going CPE integration: Further investigation of the applicability PE and P integration Forthcoming thorough evaluation with lab testing for the greenfield model 1 year from now for the target model to deliver the full interworking with the initial platforms CPE integration: Further investigation of the applicability PE and P integration Forthcoming thorough evaluation with lab testing for the greenfield model 1 year from now for the target model to deliver the full interworking with the initial platforms

23. Summary Carrier Ethernet continuity as an absolute requirement for Colt L2 products Fully approved integrated solution in 1 year time Infrastructure ready for product evolution e.g. blending of L2 & L3 services with collapsed PE Next steps: progress research on L1/L2/L3 integration