1 / 17

Network-wide Decision Making: Toward a Wafer-thin Control Plane

Network-wide Decision Making: Toward a Wafer-thin Control Plane. Jennifer Rexford, Albert Greenberg, Gisli Hjalmtysson ATT Labs Research David A. Maltz, Andy Myers, Geoffrey Xie, Jibin Zhan, Hui Zhang Carnegie Mellon University. A Well-Studied Architecture Question.

jon
Download Presentation

Network-wide Decision Making: Toward a Wafer-thin Control Plane

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Network-wide Decision Making:Toward a Wafer-thin Control Plane Jennifer Rexford, Albert Greenberg, Gisli Hjalmtysson ATT Labs Research David A. Maltz, Andy Myers, Geoffrey Xie, Jibin Zhan, Hui Zhang Carnegie Mellon University

  2. A Well-Studied Architecture Question • Smart hosts, dumb network • Network moves IP packets between hosts • Services implemented on hosts • Keep state at the edges Edge Network à IP! Edge à IP! How to partition function vertically?

  3. Shell scripts • Management Plane • Figure out what is happening in network • Decide how to change it Traffic Eng Planning tools Databases Configs SNMP netflow modems • Control Plane • Multiple routing processes on each router • Each router with different configuration program • Huge number of control knobs: metrics, ACLs, policy OSPF Link metrics Routing policies OSPF OSPF OSPF BGP BGP BGP FIB FIB FIB Inside a Single Network • Data Plane • Distributed routers • Forwarding, filtering, queueing • Based on FIB or labels Packet filters

  4. Shell scripts • Management Plane • Figure out what is happening in network • Decide how to change it Traffic Eng Planning tools Databases Configs SNMP netflow modems • Control Plane • Multiple routing processes on each router • Each router with different configuration program • Huge number of control knobs: metrics, ACLs, policy OSPF Link metrics Routing policies OSPF OSPF OSPF BGP BGP BGP FIB FIB FIB Inside a Single Network • Data Plane • Distributed routers • Forwarding, filtering, queueing • Based on FIB or labels • State everywhere! • Dynamic state in FIBs • Configured state in settings, policies, packet filters • Programmed state in magic constants, timers • Many dependencies between bits of state • State updated in uncoordinated, decentralized way! Packet filters

  5. Example – Traffic Engineering - 1 Management Plane • Routers make uncoordinated changes to their routes • Poor stability, traffic thrashing • Network-wide view needed for a network-wide goal • Load sensitive routing • OSPF distributes load info • Paths computed to avoid hotspots Control Plane Data Plane

  6. Example – Traffic Engineering - 2 Management Plane • Route planning • Learn topology • Estimate traffic matrix • Compute OSPF weights • Reconfigure routers • Must predict & undo effects of control plane • Must translate solution into settings of control plane knobs • Need ability to express desired solution OSPF Control Plane Load info Data Plane

  7. An Architecture Question to Study • How should the functionality that controls a network be divided up? • Important: everyone hates net outages • Practical: solutions can be implemented without changing IP or end-hosts • Relevant: trends toward separating decision-making from forwarding • Unsolved: problem is not solved by running BGP/OSPF on faster servers

  8. Our Proposal:Dissemination and Decision Planes • What functions require a view of entire network and network objectives? • Path selection and traffic engineering • Reachability control and VPNs • !Decision plane • What functions must be on every router to support creation of a network-wide view? • Topology discovery • Report measurements, status, resources • Install state (e.g., FIBs, ACLs) into data-plane • !Dissemination plane

  9. Good Abstractions Reduce Complexity • All decision making logic lifted out of control plane • Eliminates duplicate logic in management plane • Dissemination plane provides a control channel to/from data plane Management Plane Configs Decision Plane Control Plane FIBs, ACLs FIBs, ACLs Dissemination Data Plane Data Plane

  10. Many Implementations Possible • Decision Plane • Centralized, or • Distributed • Dissemination Plane • In-band, or • Out-of-band • Choice based on reliability requirements • Data plane evolution should be driven by needs of decision and dissemination planes

  11. Example – Traffic Engineering Reprise Decision Plane Path Computation • Network-wide view provided by Dissemination Plane • All policy, goals, decision logic located in Decision Plane • Consistent network-wide solution constructed • Decision plane can directly express desired solution Traffic Matrix Topology FIBs Dissemination Plane Load info Data Plane

  12. Example – Traffic Isolation Management Plane • Reachability control • Create routing design • Configure routing protocols • Add packet filters to patch holes where needed • Prevent some hosts/apps from communicating with others • Routing policy is very coarse grained • Packet filters are very expensive in the data plane • Missing filters can allow packets to leak, violating isolation Control Plane Route attrs Data Plane

  13. Example – Traffic Isolation Reprise Reachability matrix Decision Plane • Reachability matrix directly expresses intended goal • Path computation can jointly optimize traffic load and obey reachability constraints • Packet filters installed only where needed, and changed whenever routing changes Path Computation Traffic Matrix Topology FIBs, ACLs Dissemination Plane Load info Data Plane

  14. Challenges • Scalability for a single network • Back-of-the-envelope calculations ! no show-stoppers • Responsiveness • Reacting to unplanned failure takes an extra 40-100ms; OSPF/iBGP reconvergence today measured in seconds • Preplanning for failure is easier with network-wide view • Coordination • Minimize by having single active decision engine • Leverage distributed computing work

  15. Challenges • Dissemination plane robustness • Must survive failures of links, but be less complicated than the routing protocols it tries to replace • Hierarchy • How do two Decision Planes inter-network? • How is the boundary of the Dissemination Plane defined? 14.25

  16. Related Work • Separation of forwarding elements and control elements • IETF: FORCES, GSMP, GMPLS • SoftRouter [Lakshman] • Driving network operation from network-wide views • Traffic Engineering, Traffic Matrix computation • Centralization of decision making logic • RCP [Feamster], PCE [Farrel] • SS7 [Ma Bell] 14.75

  17. Summary • How to partition functionality inside the systems that control a network? • Dissemination and Decision Planes • Power of solution comes from: • Locating all decision making in one plane • Providing that plane with network-wide views • Directly write forwarding state to data plane • Benefits • Network-wide views • Focus on network issues, less on distributed protocols • Coordinated state updates ! better reliability 15

More Related