http://www.openflow.org/wk/index.php/openflow_tutorial 2
Tutorial Flow Section 5.1 of OpenFlowTutorial: http://www.openflow.org/wk/index.php/openflow_tutorial 3
Tutorial Setup sudomn--toposingle,3 --mac--switch ovsk--controller remote c0 Controller port6633 OpenFlow Tutorial 3hosts-1switch topology loopback (127.0.0.1:6633) s1 OpenFlow Switch s1-eth0 s1-eth1 s1-eth2 loopback (127.0.0.1:6634) dpctl (user space process) h1-eth0 h2-eth0 h3-eth0 h1 10.0.0.1 h2 10.0.0.2 h3 10.0.0.3 virtual hosts 4
Next Wed Teaching Assistant will be available during practical session hours (13:00-16:00) to help with homework questions
Part II: More SDN Apps 6
SDN application spectrum 1. Multi-tenant virtualized data centers Public and private clouds 2. WANs Google WAN Public WANs (?) 3. Enterprise networks Greater control, fewer middleboxes 7
SDN application spectrum (2) 4. Home networks Outsourced management 5. Cellular Networks Separation of service from physical infrastructure 6. Research and Education Networks National backbones College campus networks 8
OpenFlow/SDN deployment in Google WAN Urs Hölzle(Google), ONS 2012 S. Jain et al. B4: experience with a globally-deployed software defined WAN, SIGCOMM 13
openflow.org/videos 10
OpenFlow Demonstration Overview Demos FlowVisor: Network Virtualization PlugNServe: Load Balancing ElasticTree: Energy Savings Dynamic Flow Aggregation Packet and Circuit Network Convergence Using all wireless nets around us. 11
Plug-n-Serve: Load-Balancing Web Traffic using OpenFlow Goal: Load-balancing requests in unstructured networks What we are showing OpenFlow-based distributed load-balancer Smart load-balancing based on network and server load Allows incremental deployment of additional resources OpenFlow means Complete control over traffic within the network Visibility into network conditions Ability to use existing commodity hardware demo credits: N. Handigol, S. Seetharaman, M. Flajslik, N. McKeown, R. Johari 12
ElasticTree: Reducing Energy in Data Center Networks Shuts off links and switches to reduce data center power Choice of optimizers to balance power, fault tolerance, and BW OpenFlow provides network routes and port statistics The demo: Hardware-based 16-node Fat Tree Your choice of traffic pattern, bandwidth, optimization strategy Graph shows live power and latency variation demo credits: Brandon Heller, Srini Seetharaman, Yiannis Yiakoumis, David Underhill 13
Dynamic Flow Aggregation on an OpenFlow Network Scope Different Networks want different flow granularity (ISP, Backbone, ) Switch resources are limited (flow entries, memory) Network management is hard Current Solutions : MPLS, IP aggregation How OpenFlow Helps? Dynamically define flow granularity by wildcarding arbitrary header fields Granularity is on the switch flow entries, no packet rewrite or encapsulation Create meaningful bundles and manage them using your own software (reroute, monitor) Higher Flexibility, Better Control, Easier Management, Experimentation demo credits: Yiannis Yiakoumis, Jad Naous, Guido Appenzeller 14
Part III : SDN at the inter-domain level Outsourcing the Routing Control Logic: Better Internet Routing based on SDN Principles, ACM HotNets, 2012 15
Present Status Quo: Border Gateway Protocol (BGP) AS 1 ibgp ebgp Autonomous Systems (AS): Managed by one entity. Unique AS number. ebgp ibgp Exterior Gateway Protocol: BGP (ebgp and ibgp) ebgp AS 4 Internal Router Border Router AS 2 ebgp AS 3 16
Present Status Quo: BGP(2) Default inter-domain routing protocol of the Internet (path vector) Used between approximately 45 thousands Autonomous Systems (ASes) Policy playground Internet topology visualization 17
BGP has some major problems Security, Authentication (e.g., Prefix Hijacking Attacks) Delayed Routing Convergence time Policy disputes Scalability, Churn Route Oscillation, Flapping 18
BGP (and Internet Architecture) Ossicification Cannot change the core components of the Internet, like the Internet Protocol (IP) and the Internet routing system Evolving to IPv6 has taken 18 years of intense efforts The Internet routing system is likely the hardest to evolve 19
Routing Management and Optimization is Complex Diverse objectives Over-the-top service guarantees SLAs with client networks Transit cost reduction Resiliency Scalability Security Green.. Complex multi-objective optimization optimal routing Compile into low-level configuration 20
Routing is NOT core business anymore for many Internet organizations $ / Mbps Does it pay? Not always E.g.: profits in pure transit drop ($/Mbps)* Bit pipe ISP model under revision Pressure for reduced costs (OPEX) Higher-margin services: IPTV, VoIP, cloud-hosting *http://drpeering.net/white-papers/internet-transit-pricing-historical-and-projected.php 21 year
The case for Outsourcing Well-known practice to reduce-streamline OPEX Benefits from economy of scale Ecosystem of managed networking services, e.g., IBM outsources network management to AT&T. Outsourcing makes sense for Internet routing: Internet routing and optimization is hard Gets harder as the service requirements grow Large effort Small payoff Many networks only do trivial hot potato routing Idea: Routing Logic Outsourcing 22
Outsourcing the Routing Logic OUTSOURCING SERVICE CONTRACTOR ISP High-level services ROUTING INFRASTRUCTURE Configuration Control Logic SLA High-level services ISP ROUTING INFRASTRUCTURE Focus on profitable services on top of routing Buy expertise from specialized contractor Form interactive business relationship 23
Outsourcing: smooth transition Transition stages: Consultation, Information about Best Practices Routing Control Logic Path Computation Low-level control of FIBs, RIBs Routing Control Plane During the transition we: Shape our own policies (based on business model) Express our requirements to the contractor Policy privacy* based on trusted party model, NDAs If not satisfied backtrack 24 *No leakage to competitors
SDN: enabling simpler outsourcing Routing App #1 Routing App #2 Single-Domain Routing OS Platform Contractor Policy Requirements, Network State, Monitoring Info, Routing Configuration Packet Forwarding Packet Forwarding Packet Forwarding Packet Forwarding Packet Forwarding Client Domain 25
New multi-domain services Thinking bigger: cumulative outsourcing Optimizations Troubleshooting Programmable Transit Multi-Domain Routing OS Platform Contractor Domain A Cumulative Information Exchange Packet Forwarding Domain B Packet Forwarding Packet Forwarding Packet Forwarding Packet Forwarding 26
Additional benefit: legacycompatible evolution CONTRACTOR #1 CONTRACTOR #2 CLIENT AS CLIENT AS CLIENT AS CLIENT AS BGP BGP Ecosystem of outsourcing service contractors - clients New routing-signaling protocols within the clusters New protocols for contractor interoperability Legacy Compatibility(BGP) INDEPENDENT AS 27 BGP CLIENT AS
Candidate Contractors Larger and tier-1 ISPs (but: potential conflicts) External specialized parties (more objective) Why? Considerable expertise in routing Incentive for a new service type provision (outsourcing) Opportunity for an economy of scale Example: AT&T Tier-1 ISP Market leader in handling outsourced network services 28
Candidate Clients Small or medium sized network providers Why? Global trend: Higher and higher interconnectivity Flattening of the AS topology graph* Need for sophisticated Traffic Engineering Complexity increases Who should handle the complexity? Observation: Large number of potential clients out there (~10s of thousands) 29 *C. Labovitzet al. Internet Inter-Domain Traffic. SIGCOMM 2010.
Contractor: Summary of incentives Offer better service to clients (efficiency, stability, availability) Respect and build upon their requirements Achieve economy of scale Client: Invest in high-margin services Outsource a significant portion of complexity Interplay between entities: Incremental Optimizations in Routing (inter-as level) New opportunities-services: programmable transit, global IPTV 30
Sample Architecture (single client contractor) SERVICE CONTRACTOR ROUTING OS INTRA-/INTER- AS ROUTING CONTROL LOGIC POLICIES NETWORK STATE MONITORING DATA BGP MESSAGES SECURE COMMUNICATION ROUTING CONFIGURATION BGP MESSAGES BGP CONTRACTOR S LOCAL DELEGATE MONITORING (SNMP, NETFLOW, ETC) CONTROL (OpenFlow, NETCONF, CLIs) CLIENT DOMAIN 31
Research Goals -Challenges 32
Framework: Multi-Domain Network High-Level goals Operating System Provision of outsourced routing services across domains Based on Outsourcing and Logical Centralization Model How to achieve Define proper south, north and east-west APIs Description, design and implementation of a multidomain NOS, oriented to routing and path computation System properties Performance, Service Orchestration, Scalability, Resiliency, Security, Privacy Autonomy, Evolvability, Interoperability Implications Centralization vs Distribution, Fail-over Tussles 33
Algorithms Applications - High-Level goal Services Formation and analysis of new inter-domain routing algorithms, applications and services How to achieve Use of multi-domain NOS framework Define proper northbound interfaces Outsourcing contractor = mediator SLAs Examples Routing optimization algorithms Collaborative security Cooperative troubleshooting Policy conflict resolution 34
Recap: Routing Outsourcing CONTRACTOR #1 CONTRACTOR #2 CLUSTER OF CLIENTS CLUSTER OF CLIENTS BGP LEGACY DOMAINS BGP Benefits Legacy-compatible inter-domain control plane evolution Change BGP within the cluster Inter-domain optimizations Multi-domain traffic engineering Economy of Scale Challenges (many more ) Resiliency/scalability of multidomain routing control platform Evaluate viability of routing outsourcing business model Incentive-based multi-domain optimizations 35
Further Reading Outsourcing the Routing Control Logic: Better Internet Routing based on SDN Principles Vasileios Kotronis, Xenofontas Dimitropoulos and Bernhard Ager ACM HotNets, October 2012. Outsourcing Routing using SDN: The Case for a Multi-Domain Routing Operating System Vasileios Kotronis, Xenofontas Dimitropoulos and Bernhard Ager Open Networking Summit 2013 (poster), April 2013. 36
Thank you! 37