Transcript Document
Optimizing Metro Ethernet Tim Hubbard Nortel Networks THIS IS RHK Telecom Negotiation Tips Educating the Enterprise Customer • “Aggregate spending into one contract to draw more bids” – Support for Triple play of Voice, Video, and Data • “Include renegotiation clause in the event of a downturn” – Dynamic bandwidth provisioning • “Assure latest technology will be used” – Ethernet-based services in high demand • “Include termination clauses for poor service” – Carrier-class availability and verifiable SLAs THIS IS Metro Service Platform • Advanced Traffic Management – Bi-directional bandwidth control – Hierarchical quality of service – High performance hardware implementation • Global Scalability – VPLS – H-VPLS • Carrier-class Resiliency THIS IS Traffic Management THIS IS Differentiated Services Requirements • Flexible provisioning of Bandwidth: – Bandwidth control: Rate shaping, policing • Differentiated service delivery: – Proven end-to-end Quality of Service • Wire-speed Implementation of Traffic management – Low latency Subscriber Services Central Office Service Provisioning SLA Monitoring THIS IS Traffic Management Illustration Incoming Traffic Measure by Classification (Coloring by Apps) Data Video Voice Policing and Remarking Discard or re-marked as Best Effort (Tired ingress Control) Best Effort Priority mapped into .1p tag in vMAN header Low Priority Mid Priority High Priority To Switch Fabric THIS IS Traffic Management Rate Shaping • Gigabit throughput on any port with over-subscription – Flexible provisioning of bandwidth per subscriber • Bi-Directional Rate shaping – Support for ingress and egress rate shaping • 3 Color Rate Shaping • Committed Information Rate (CIR) for guaranteed traffic • Peak Rate (PR) for Burst Services • Marking for non-conforming excess traffic • Granular and Accurate Rate Shaping with 1 and 2 kbps increments for Ingress and Egress rate limiting; • Statistics collection: CIR, PR and Dropped Bytes • Reserve or limit bandwidth per application/per subscriber THIS IS Traffic Management Quality of Service • Quality of Service Implementation: – Bandwidth allocation per user / per application – Classification • “QoS” aware and unaware applications – Marking • DiffServ – IP QoS, 802.1p – Ethernet QoS • Classify, Mark/Re-mark, Schedule – Scheduling • 8 hardware-based queues per port to support: THIS IS Traffic Management 1st level 2nd 1st level 2nd level 3rd level Per Service Per Customer Per Physical Port (VLAN) (VMAN) level Per Service Per Customer (VLAN) (VMAN) Cust A Cust A Cust B Cust Z BW Dynamic Allocation = Enables customer to use all bandwidth and fill it in order of service priority Each customer has the minimum (=CIR), and maximum (=PR) bandwidth setting Available resources within the customer’s total bandwidth are allocated to the other service types in a weighted fashion after priority CIR is met THIS IS High Performance Traffic Management Supports triple-play: voice, video and data • Wire-speed implementation of network processing and control • Non-blocking architecture delivering “zero” latency • No impact on performance when features are activated: – Rate shaping – QoS Maximum Latency For Voice • Latency is independent of network load THIS IS High Performance Traffic Management Supports triple-play: voice, video and data • Scalable Multicast Service Delivery – Wire-Speed forwarding of Unicast and Multicast simultaneously • Adding multicast will not affect unicast forwarding – Supports increasing number of user • Every new receiver will not slow forwarding • Every port in the network can potentially be a receiver – Supports increasing number of video streams • Every new transmitter will not slow forwarding • Every port in the network can potentially be a transmitter – No performance drop for video applications • Video Conferencing • Video Streaming and distribution THIS IS Global Scalability THIS IS Global Scalability Layer 2 Multiplexing • QinQ and MinM • End-to-end Layer 2 subscriber services • Provides subscriber scalability and separation of subscriber/provider Ethernet control traffic • Traditional Layer 2 VLAN service in access • Preservation of customer VLAN tags Subscirber Layer-2 VLAN VMANTunnel or QinQ tunnel VMAN Customer Network Ethernet VMAN Tunneled Core Customer Network Customer Network VMAN to VMAN Switching Customer Network THIS IS Global Scalability Hybrid Ethernet/VPLS Metro EAPS (RFC 3619) (Ethernet Automatic Protection System) Ethernet (vMAN) Access Ring Ethernet (vMAN) Access Ring Highly-scalable Metro Core (VPLS Switched / IP Routed) MPLS L2VPN Tunnels (For Differentiated Service Classes & Fast Path Restoration) IP Networks (Internet) & Content Providers Either Ring or Mesh Architecture THIS IS “Carrier-Class” Resiliency THIS IS Availability Multi-layer Network Resiliency Layer 1 – Physical Layer 2 – Ethernet Topologies Layer 3 – Routing Software Redundant Port Chassis Blade Port E- BGP Internet 48si Physical Redundancy • Port, blade, chassis • Redundant Power supply • Redundant Switching Fabric • Hot-swappable • Hitless upgrades, patching, etc • Protected memory VRRP Link Aggregation OSPF Ring • • • • STP RSTP EAPS Proprietary VLAN Aggregation • Virtualisation of software threads and processing • Software sparing THIS IS Availability Management / Fabric Resiliency Hitless Upgrade Hitless Failover Switches maintain state and forwarding capabilities even in the event of software or hardware failures Existing Core Software and firmware can be updated without taking switch out of service Hardware protection Metro Ethernet Network Protects against cable, port and I/O module failure THIS IS Availability, Solutions for Layer-2 Resiliency • IEEE 802.1w/s IEEE 802.1D – Standard - Rapid spanning tree and Multiple instance spanning tree • IEEE 802.3ad – Link aggregation often used as 1:1 protection switching – Provides the benefit of enabling use of “sparing” or standby link when both interfaces are operating normally • Dual homed software based redundant ports; • Virtual Router Redundancy Protocol (VRRP): – Topology intelligent Ethernet protection; • Ethernet Automatic Protection System (EAPS) RFC 3619 – – – – – Ethernet ring-based protection Multiple domains on a ring Multiple domains on a node VLAN can be member of multiple domains Co-exists with STP THIS IS Availability – EAPS v2 Ethernet Automatic Protection Switching • Layer 2 Fast protection switching for Ethernet ring topologies • EAPS - Informational RFC 3619 • Fail-over < 50 Milliseconds – Faster than Fast STP in a ring: – Fail-over independent of number of nodes in the ring • Traffic flow in both directions paths selectable per VLAN – Enables Traffic Engineering • Supports simple and complex ring topologies – Dual attached rings – Subtended rings THIS IS Metro Service Summary • Advanced Traffic Management – Bi-directional bandwidth control – Hierarchical quality of service – High performance hardware implementation • Global Scalability – QinQ at the Edge – H-VPLS into the core • Carrier-class Resiliency – Ethernet APS – RPR – EoSonet/SDH THIS IS End THIS IS