dynamic network fabric for nfv
play

Dynamic Network Fabric for NFV Mario A. Snchez , Joon-Myung Kang, - PowerPoint PPT Presentation

Dynamic Network Fabric for NFV Mario A. Snchez , Joon-Myung Kang, Ying Zhang Research Scientist, Networking, IoT and Mobility Lab Overlay Virtual Network Fabrics Overlay virtual network fabrics are essential in the evolution and deployment


  1. Dynamic Network Fabric for NFV Mario A. Sánchez , Joon-Myung Kang, Ying Zhang Research Scientist, Networking, IoT and Mobility Lab

  2. Overlay Virtual Network Fabrics Overlay virtual network fabrics are essential in the evolution and deployment >> Potential Functional Egress Potential Egress << of NFV and distributed Service Provider datacenters Secured-Peering 3 SDN based Overlays on various underlay Secured- networks providing datacenter/cloud Peering 2 network virtualization Secured- Peering 4 Potential Egress << << Ingress traffic Secured-Peering 5 Secured-Peering 1 Confidential

  3. Benefits - Enhanced Smart Virtual Network Fabric • Higher service capacity with smaller resource footprint • Significant reduction in resource footprint and operational/mgmt overhead while maintaining SLAs • Applicable to enhance multiple NFV solutions with multi-site deployments: vCPE,vIMS, vEPC Confidential

  4. Enhanced Smart Virtual Network Fabrics for NFV/SPs • Intelligent workload aware overlay over IP networks with real-time measurements and predictive analytics SLA Monitor and Manager • Monitor end-to-end SLA per-flow or flow aggregates with Monitors Cloudified NFV solution paths over underlay and SFCs • Workload driven smart dynamic solution for distributed Bandwidth-Aware Underlay-Aware datacenters/POPs Resilient VNF Overlay Placement Optimizes the Optimizes the Network Environment Compute Environment HPE NFV Solution E.g., NFV System, ConteXtream Base HPE vE-CPE Foundation Confidential

  5. Virtual Network Fabrics Overview vE-CPE vE-CPE Cloudified vE-CPE vE-CPE Challenges • Increased agility – faster time to revenue • Over-provisioning resources per enterprise branch – Replace per branch dev-ops with continual, decoupled dev-ops inefficient - increases CAPEX per enterprise, applied to carrier cloud as whole • Slow and Complex semi-manual dev-ops with risky • Lower TCO onboarding increases OPEX, slows time to revenue – CAPEX: Reduces vE-CPE resources (upto 4x in scenarios) • POP or rack failures require 1+1 availability – OPEX: Simplifies mgmt. and provisioning • Inability to distribute demand spikes to POP overlay • Improved Reliability and Availability edges without protection from SLA violations – Distributed POPs are resource pools with increased availability • Inefficient provisioning of VNF resource mgmt. leads – Eliminate impact of POP/Rack failures, spikes, and stampedes to fragmentation of resources higher TCO • SLAs monitored to guarantee user experience while traffic dynamically mapped for maximum efficiency Confidential

  6. Multi-site Topology POP4 172.24.4.14 172.24.4.15 POP5 172.24.4.114 POP2 6 9 172.24.4.115 5 8 172.24.4.112 2 7 4 172.24.4.12 1 3 POP1 172.24.4.11 172.24.4.13 172.24.4.111 Confidential

  7. Interactions between different components Use cases • A customer requires a new chain: • 1. (PoPx = local PoP)->2->3->4->1- 5. Is chain x set up >2…5->6 Control correctly? Does chain x • A rack fails: satisfy SLA? • 3->4->1->2->5->6 • Traffic spike • 6 ->3->4->1->2 4. PoP x 2. Not enough 1. Does PoP x have resource or sufficient resource? If Chain is created so, create the chain successfully 3 . Which remote 6. Chain x violates PoP to use? policy/SLA VNF SLA UAO Placement Verifier Confidential

  8. Underlay-Aware Overlay (UAO) • Utilizes knowledge of underlay network performance to optimize and improve overlay performance – Dynamically maps traffic optimal resources – Granular support: per-flow overlay route decision logic – Standards-based: Plug-in to ODL overlay control • UAO Routing Optimization – Aware of joint paths – Aware of load injected from paths with joint bottlenecks – Distribute local load to equalize load on bottlenecks Max Delay assuming stable peer routing policies 6000 • Increases reliability and availability 5000 – Eliminates impact of rack or PoP failures 4000 – Increases efficiency of underlay to absorb failures 3000 2000 1000 0 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2 2.1 2.2 2.3 2.4 2.5 2.6 Confidential

  9. Underlay Aware Overlay POPA à POPB POPA à POPC POPB à POPA POPC à POPA POPB à POPC POPC à POPB Confidential

  10. Underlay Aware Overlay POPB à POPA POPB à POPC POPA à POPB POPA à POPC POPC à POPA POPC à POPB POPB à POPA POPB à POPA POPB à POPC POPB à POPC POPA à POPB POPA à POPB POPA à POPC POPA à POPC POPC à POPA POPC à POPA POPC à POPB POPC à POPB Confidential

  11. POPB à POPA 0.9 Underlay Aware Overlay POPB à POPC 0.1 POPA à POPB 0.8 POPA à POPC 0.7 POPC à POPA 0.1 POPC à POPB 0.3 POPB à POPA 0.9 POPB à POPC 0.1 POPB à POPA 0.9 POPA à POPB 0.8 POPB à POPC 0.1 POPA à POPC 0.7 POPA à POPB 0.8 POPC à POPA 0.1 POPA à POPC 0.7 POPC à POPB 0.3 POPC à POPA 0.1 POPC à POPB 0.3 Confidential

  12. Path Correlation Matrix Confidential

  13. Smart Virtual Network Fabric Overview Confidential

  14. Underlay-Aware Overlay (UAO) • Utilizes knowledge of underlay network performance to optimize and improve overlay performance – Dynamically maps traffic optimal resources – Granular support: per-flow overlay route decision logic – Standards-based: Plug-in to ODL overlay control • UAO Routing Optimization – Aware of joint paths – Aware of load injected from paths with joint bottlenecks – Distribute local load to equalize load on bottlenecks Max Delay assuming stable peer routing policies 6000 • Increases reliability and availability 5000 – Eliminates impact of rack or PoP failures 4000 – Increases efficiency of underlay to absorb failures 3000 2000 1000 0 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2 2.1 2.2 2.3 2.4 2.5 2.6 Confidential

  15. Thank You

Recommend


More recommend