1 Rethinking Network Control and Management David A. Maltz

Slides:



Advertisements
Similar presentations
David A. Maltz Carnegie Mellon University/Microsoft Research
Advertisements

New Directions in Enterprise Network Management Aditya Akella University of Wisconsin, Madison MSR Networking Summit June 2006.
Path Splicing with Network Slicing
Path Splicing with Network Slicing Nick Feamster Murtaza Motiwala Santosh Vempala.
MPLS VPN.
Generalized Multiprotocol Label Switching: An Overview of Signaling Enhancements and Recovery Techniques IEEE Communications Magazine July 2001.
Logically Centralized Control Class 2. Types of Networks ISP Networks – Entity only owns the switches – Throughput: 100GB-10TB – Heterogeneous devices:
Deployment of MPLS VPN in Large ISP Networks
IEEE HPSR IP Network Background and Strategy Milestones  Started as a Internet backbone/IGW  Expansion with MAN networks  Tripleplay and multimedia,
1 EL736 Communications Networks II: Design and Algorithms Class3: Network Design Modeling Yong Liu 09/19/2007.
Multi-Layer Switching Layers 1, 2, and 3. Cisco Hierarchical Model Access Layer –Workgroup –Access layer aggregation and L3/L4 services Distribution Layer.
Consensus Routing: The Internet as a Distributed System John P. John, Ethan Katz-Bassett, Arvind Krishnamurthy, and Thomas Anderson Presented.
15-744: Computer Networking
1 In VINI Veritas: Realistic and Controlled Network Experimentation Jennifer Rexford with Andy Bavier, Nick Feamster, Mark Huang, and Larry Peterson
1 Route Control Platform Making the Network Act Like One Big Router Jennifer Rexford Princeton University
Traffic Engineering Jennifer Rexford Advanced Computer Networks Tuesdays/Thursdays 1:30pm-2:50pm.
Shadow Configurations: A Network Management Primitive Richard Alimi, Ye Wang, Y. Richard Yang Laboratory of Networked Systems Yale University.
A Routing Control Platform for Managing IP Networks Jennifer Rexford Computer Science Department Princeton University
1 Problems and Solutions in Enterprise Network Control: Motivations for a 4D Architecture David A. Maltz Microsoft Research Joint work with Albert Greenberg,
Dynamics of Hot-Potato Routing in IP Networks Renata Teixeira (UC San Diego) with Aman Shaikh (AT&T), Tim Griffin(Intel),
Shadow Configurations: A Network Management Primitive Richard Alimi, Ye Wang, and Y. Richard Yang Laboratory of Networked Systems Yale University February.
1 Design and implementation of a Routing Control Platform Matthew Caesar, Donald Caldwell, Nick Feamster, Jennifer Rexford, Aman Shaikh, Jacobus van der.
A Routing Control Platform for Managing IP Networks Jennifer Rexford Princeton University
Internet Routing (COS 598A) Today: Telling Routers What to Do Jennifer Rexford Tuesdays/Thursdays.
Network Monitoring for Internet Traffic Engineering Jennifer Rexford AT&T Labs – Research Florham Park, NJ 07932
A Routing Control Platform for Managing IP Networks Jennifer Rexford Princeton University
Tesseract A 4D Network Control Plane
A Routing Control Platform for Managing IP Networks Jennifer Rexford Princeton University
A General approach to MPLS Path Protection using Segments Ashish Gupta Ashish Gupta.
Institute of Technology, Sligo Dept of Computing Semester 3, version Semester 3 Chapter 3 VLANs.
1 Network-wide Decision Making: Toward a Wafer-thin Control Plane Jennifer Rexford, Albert Greenberg, Gisli Hjalmtysson ATT Labs Research David A. Maltz,
Jennifer Rexford Princeton University MW 11:00am-12:20pm Wide-Area Traffic Management COS 597E: Software Defined Networking.
Chapter 1: Hierarchical Network Design
OpenFlow-Based Server Load Balancing GoneWild Author : Richard Wang, Dana Butnariu, Jennifer Rexford Publisher : Hot-ICE'11 Proceedings of the 11th USENIX.
1 Computer Communication & Networks Lecture 22 Network Layer: Delivery, Forwarding, Routing (contd.)
Software Defined-Networking. Network Policies Access control: reachability – Alice can not send packets to Bob Application classification – Place video.
Happy Network Administrators  Happy Packets  Happy Users WIRED Position Statement Aman Shaikh AT&T Labs – Research October 16,
Reducing Transient Disconnectivity using Anomaly-Cognizant Forwarding Andrey Ermolinskiy, Scott Shenker University of California – Berkeley and ICSI.
Jennifer Rexford Fall 2014 (TTh 3:00-4:20 in CS 105) COS 561: Advanced Computer Networks BGP.
1 Rethinking Network Control & Management The Case for a New 4D Architecture David A. Maltz Carnegie Mellon University Joint work with Albert Greenberg,
Central Control over Distributed Routing fibbing.net SIGCOMM Stefano Vissicchio 18th August 2015 UCLouvain Joint work with O. Tilmans (UCLouvain), L. Vanbever.
Sub-ip - 1 Blurring the Lines Between Circuits and Protocols: Plans to Re-Organize Sub-IP Technologies in the IETF Scott Bradner Harvard University.
11 Tesseract* A 4D Network Control Plane Hong Yan, David A. Maltz, T. S. Eugene Ng Hemant Gogineni, Hui Zhang, Zheng Cai *Tesseract is a 4-dimensional.
A Firewall for Routers: Protecting Against Routing Misbehavior1 June 26, A Firewall for Routers: Protecting Against Routing Misbehavior Jia Wang.
A Snapshot on MPLS Reliability Features Ping Pan March, 2002.
SDN AND OPENFLOW SPECIFICATION SPEAKER: HSUAN-LING WENG DATE: 2014/11/18.
Intradomain Traffic Engineering By Behzad Akbari These slides are based in part upon slides of J. Rexford (Princeton university)
1 Routing Design in Operational Networks: A Look from the Inside David A. Maltz, Geoffrey Xie, Jibin Zhan, Hui Zhang Carnegie Mellon University Gisli Hjalmtysson,
1 Computer Communication & Networks Lecture 21 Network Layer: Delivery, Forwarding, Routing Waleed.
Evolving Toward a Self-Managing Network Jennifer Rexford Princeton University
High-Speed Policy-Based Packet Forwarding Using Efficient Multi-dimensional Range Matching Lakshman and Stiliadis ACM SIGCOMM 98.
Evolving Toward a Self-Managing Network Jennifer Rexford Princeton University
A Snapshot on MPLS Reliability Features Ping Pan March, 2002.
Separating Routing From Routers Jennifer Rexford Princeton University
Fabric: A Retrospective on Evolving SDN Presented by: Tarek Elgamal.
Separating Routing From Routers Jennifer Rexford Princeton University
Preliminaries: EE807 Software-defined Networked Computing KyoungSoo Park Department of Electrical Engineering KAIST.
BUFFALO: Bloom Filter Forwarding Architecture for Large Organizations Minlan Yu Princeton University Joint work with Alex Fabrikant,
Multi Node Label Routing – A layer 2.5 routing protocol
Shadow Configurations: A Network Management Primitive
15-744: Computer Networking
15-744: Computer Networking
Routing Jennifer Rexford.
ETHANE: TAKING CONTROL OF THE ENTERPRISE
6.829 Lecture 13: Software Defined Networking
Introduction to Internet Routing
Intra-Domain Routing Jacob Strauss September 14, 2006.
COS 561: Advanced Computer Networks
COS 461: Computer Networks
Control-Data Plane Separation
Presentation transcript:

1 Rethinking Network Control and Management David A. Maltz

2 22 Context for Network Control and Management  Many different network environments  Access, backbone networks  Data-center networks, enterprise/campus  Many different technologies  Longest-prefix routing, label switching, circuit switching  IP, Ethernet, MPLS, optical circuits  Outsourcing of responsibility into the network  Middle-boxes: firewalls, network monitoring, …  Many different policies  Routing, reachability, transit, traffic engineering, robustness

3 33 ATT/CMU Study of 31 Production networks  Provider & enterprise networks ( routers)  Many different routing designs  Packet filters, multiple OSPF instances, multiple ASs Router ID 8810 Lines in config file

4 44 Fundamental Problem: Wrong Abstractions Management Plane Figure out what is happening in network Decide how to change it Shell scriptsTraffic Eng Databases Planning tools OSPF SNMP netflowmodems Configs OSPF BGP Link metrics OSPF BGP OSPF BGP Control Plane Multiple routing processes on each router Each router with different configuration program Huge number of control knobs: metrics, ACLs, policy FIB Routing policies Packet filters Data Plane Distributed routers Forwarding, filtering, queueing Based on FIB or labels

5 55 Inside a Single Network  Data Plane  Distributed routers  Forwarding, filtering, queueing  Based on FIB or labels Management Plane Figure out what is happening in network Decide how to change it Shell scripts Traffic Eng Databases Planning tools OSPF SNMPnetflowmodems Configs OSPF BGP Link metrics OSPF BGP OSPF BGP Control Plane Multiple routing processes on each router Each router with different configuration program Huge number of control knobs: metrics, ACLs, policy FIB Routing policies Packet filters 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!

6 66 Inside a Single Network  Data Plane  Distributed routers  Forwarding, filtering, queueing  Based on FIB or labels Management Plane Figure out what is happening in network Decide how to change it Shell scripts Traffic Eng Databases Planning tools OSPF SNMPnetflowmodems Configs OSPF BGP Link metrics OSPF BGP OSPF BGP Control Plane Multiple routing processes on each router Each router with different configuration program Huge number of control knobs: metrics, ACLs, policy FIB Routing policies Packet filters 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! Logic everywhere! Path Computation built into routing protocols Routing Policy distributed across the routers Packet Filters placed by tools in Mng. Plane No way to arbitrate inconsistencies between logic

7 77 Control Plane: The Key Leverage Point  Great Potential: control plane determines the behavior of the network  Reaction to events, reachability, services  Great Opportunities  Each network (administrative domain) has its own control plane  A radical clean-slate control plane can be deployed –Agnostic to user data format: IPv4/v6, ethernet, circuit –No changes to end-system software  Control plane is the nexus of network evolution –Changing the control plane logic can smooth transitions in network technologies and architectures

8 88 An Alternative: The 4D Architecture  Key principles  Network-level objectives  Network-wide views  Direct control  Corollaries  Predictable behavior (including overload threshold)  Zero device-specific or manual configuration  Data plane support for network-wide view  Define objectives in terms of organizationally salient entities

9 99 Good Abstractions Reduce Complexity All decision making logic lifted out of control plane  Eliminates duplicate logic in management plane  Dissemination plane provides robust communication to/from data plane switches Management Plane Control Plane Data Plane Decision Plane Dissemination Data Plane Configs FIBs, ACLs

10 Overview of the 4D Architecture Decision Plane:  All management logic implemented on centralized servers making all decisions  Decision Elements use views to compute data plane state that meets objectives, then directly writes this state to routers Decision Dissemination Discovery Data Network-level objectives Direct control Network-wide views

11 Concerns and Challenges  Distributed Systems issues  How will communication between routers and DEs survive failures in the network?  Latency means DE’s view of network is behind reality. Will the control loop be stable?  What is the overhead to/from the DEs?  What happens in a network partition?  Networking issues  Does the 4D simplify control and management?  Can we create logic to meet multiple objectives?

12 Evaluation of the 4D Prototype  Evaluated using Emulab (  Linux PCs used as routers (650 – 800MHz)  Tested on 9 enterprise network topologies ( routers each) Example network with 49 switches and 5 DEs

13 Performance of the 4D Prototype Trivial prototype has performance comparable to well- tuned production networks  Recovers from single link failure in < 300 ms  < 1 s response considered “excellent”  Faster forwarding reconvergence possible  Survives failure of master Decision Element  New DE takes control within 1 s  No disruption unless second fault occurs  Gracefully handles complete network partitions  Less than 1.5 s of outage

14 Thanks!

15 Future Work  Scalability  Evaluate over 1-10K switches, K routes  Networks with backbone-like propagation delays  Structuring decision logic  Arbitrate among multiple, potentially competing objectives  Unify control when some logic takes longer than others  Protocol improvements  Better dissemination and discovery planes  Deployment in today’s networks  Data center, enterprise, campus, backbone (RCP)

16 Future Work  Expand relationships with security  Securing the infrastructure  Using 4D as mechanism for monitoring/quarantine  Formulate models that establish bounds of 4D  Scale, latency, stability, failure models, objectives  Generate evidence to support/refute principles

17 Themes of Network Control & Management Holistic Design  Many different technologies – a few common problems  Find the right abstractions: exploit commonality Clean Slate  How much autonomy do routers/switches need?  New principles for controlling networks  Separate networking issues from distributed system issues Leverage Network Structure  Many different types of networks exist - each with different objectives and topologies

18 Recent Publications  G. Xie, J. Zhan, D. A. Maltz, H. Zhang, A. Greenberg, G. Hjalmtysson, J. Rexford, “On Static Reachability Analysis of IP Networks,” IEEE INFOCOM 2005, Orlando, FL, March  J. Rexford, A. Greenberg, G. Hjalmtysson, D. A. Maltz, A. Myers, G. Xie, J. Zhan, H. Zhang, “Network-Wide Decision Making: Toward a Wafer-Thin Control Plane,” Proceedings of ACM HotNets-III, San Diego, CA, November  D. A. Maltz, J. Zhan, G. Xie, G. Hjalmtysson, A. Greenberg, H. Zhang, “Routing Design in Operational Networks: A Look from the Inside,” Proceedings of the 2004 Conference on Applications, Technologies, Architectures, and Protocols for Computer Communications (ACM SIGCOMM 2004), Portland, Oregon,  D. A. Maltz, J. Zhan, G. Xie, H. Zhang, G. Hjalmtysson, A. Greenberg, J. Rexford, “Structure Preserving Anonymization of Router Configuration Data,” Proceedings of ACM/Usenix Internet Measurement Conference (IMC 2004), Sicily, Italy, 2004.

19 A Clean-slate Design  What are the fundamental causes of network problems?  How to secure the network and protect the infrastructure?  What functionality needs to be distributed – what can be centralized?  How to reduce/simplify the software in networks?  What would a “RISC” router look like?  How to leverage technology trends?  CPU and link-speed growing faster than # of switches

20 Three Principles for Network Control & Management Network-level Objectives:  Express goals explicitly  Security policies, QoS, egress point selection  Do not bury goals in box-specific configuration Management Logic Reachability matrix Traffic engineering rules

21 Three Principles for Network Control & Management Network-wide Views:  Design network to provide timely, accurate info  Topology, traffic, resource limitations  Give logic the inputs it needs Management Logic Reachability matrix Traffic engineering rules Read state info

22 Three Principles for Network Control & Management Direct Control:  Allow logic to directly set forwarding state  FIB entries, packet filters, queuing parameters  Logic computes desired network state, let it implement it Management Logic Reachability matrix Traffic engineering rules Read state info Write state

23 Overview of the 4D Architecture Dissemination Plane:  Provides a robust communication channel to each router – and robustness is the only goal!  May run over same links as user data, but logically separate and independently controlled Decision Dissemination Discovery Data Network-level objectives Direct control Network-wide views

24 Overview of the 4D Architecture Discovery Plane:  Each router discovers its own resources and its local environment  E.g., the identity of its immediate neighbors Decision Dissemination Discovery Data Network-level objectives Direct control Network-wide views

25 Overview of the 4D Architecture Data Plane:  Spatially distributed routers/switches  Can deploy with today’s technology  Looking at ways to unify forwarding paradigms across technologies Decision Dissemination Discovery Data Network-level objectives Direct control Network-wide views

26 Fundamental Problem: Conflation of Issues  Ideal case: all routing information flooded to all routers inside network  Robustness achieved via flooding  Reality: routing information filtered and aggregated extensively  Route filtering used to implement security and resource policies  Route aggregation used to achieve scalability

27 4D Separates Distributed Computing Issues from Networking Issues  Distributed computing issues ! protocols and network architecture  Overhead  Resiliency  Scalability  Networking issues ! management logic  Traffic engineering and service provisioning  Egress point selection  Reachability control (VPNs)  Precomputation of backup paths

28 4D Can Leverage Network Structure  Decision plane logic can be specialized for structure of each physical network  Distributed protocols must be prepared for arbitrary topology graphs  4D enables network logic specialized differently for access and for backbone  E.g., creating aggregation tree in access network  Advantages  Faster route computations  Retain flexibility to evolve network as needed  Support transition to 100x100 architecture

29 The Feasibility of the 4D Architecture We designed and built a prototype of the 4D Architecture  4D Architecture permits many designs – prototype is a single, simple design point  Decision plane  Contains logic to simultaneously compute routes and enforce reachability matrix  Multiple Decision Elements per network, using simple election protocol to pick master  Dissemination plane  Uses source routes to direct control messages  Extremely simple & robust  Quickly route around failed data links, even multiple failures