Sweet Little Lies: Fake Topologies for Flexible Routing

Transcription

1 Sweet Little Lies: Fake Topologies for Flexible Routing Stefano Vissicchio University of Louvain HotNets 27th October 2014 Joint work with Laurent Vanbever (Princeton) and Jennifer Rexford (Princeton)

2 Fibbing

3 Fibbing control routers FIB, lying to routers

4 ??? Fibbing control routers FIB, lying to routers

5 We use lies to overcome inflexibilities of traditional networks traditional networks

6 We use lies to work around challenges of OpenFlow-like networks OpenFlow networks

7 We use lies to combine the advantages of OpenFlow and of traditional networking Fibbing

8 Operators need flexibility for intra-domain routing fine-grained traffic engineering optimize the available resources provision backup paths quickly and predictably react to failures deploy advanced services e.g., steer traffic through middleboxes

9 Consider this network where a source sends traffic to 2 destinations A B C D source destination traffic flow

10 As congestion appears on the (C,D) link, operators may want to divert the blue flow to A initial desired A B A B C D C D

11 Moving only the orange flow to A is impossible with an IGP as both destinations are connected to D initial desired A 1 B A?? B 10 1???? C 3 D C?? D impossible by reweighing the IGP links

12 Currently, operators have two ways to improve flexibility virtual circuit based solutions (MPLS) SDN based solutions (OpenFlow)

13 Both solutions comes at a significant cost virtual circuit based solutions (MPLS) control- and data-plane overhead SDN based solutions (OpenFlow) deployment costs, new challenges of a novel paradigm

14 Fibbing achieves flexible routing in an existing network, à la SDN

15 Fibbing achieves flexible routing in an existing network, à la SDN unmodified routers (checked in testbed!)

16 Fibbing achieves flexible routing in an existing network, à la SDN what does it mean?

17 The Fibbing controller program routers! (bypassing proprietary configuration languages) Controller central control of forwarding paths Control-Plane Control-Plane Control-Plane distributed forwarding path computation Data-Plane Data-Plane Data-Plane Cisco Juniper Alcatel

18 The controller uses an API that all routers understand (hint: not OpenFlow) Controller??? Control-Plane Control-Plane Control-Plane Data-Plane Data-Plane Data-Plane Cisco Juniper Alcatel

19 Link-state IGPs are actually good for something, to control router behavior messages are standardized all routers must speak the same language behaviors are well-defined and understood e.g., shortest-path routing implementations are robust and widely-deployed nearly all networks out there run OSPF or IS-IS

20 how? Fibbing achieves flexible routing in an existing network, à la SDN

21 the controller tricks IGP routers with small lies about fake nodes, links and destinations Fibbing achieves flexible routing in an existing network, à la SDN

22 As congestion appears on the (C,D) link, operators may want to divert the blue flow to A initial desired A 1 B A B 10 1 C 3 D C D

23 Fibbing can move the orange flow by adding a fake node announcing the blue destination fake node A 1 B V C 3 D

24 Fibbing can move the orange flow by adding a fake node announcing the blue destination A 1 B V new control-plane path for C 1 C 3 D

25 Fibbing can move the orange flow by adding a fake node announcing the blue destination new data-plane next-hop of C A 1 B V1 1 1 C 3 D

26 Fibbing works today! (tested with off-the-shelf Cisco routers) A 1 B V1 1 1 C 3 D

27 Is Fibbing expressive? Does Fibbing scale?

28 Is Fibbing expressive? Yes! Does Fibbing scale?

29 Fibbing is expressive Theorem Any set of forwarding DAGs can be enforced by Fibbing

30 Fibbing is expressive Theorem Any set of forwarding DAGs can be enforced by Fibbing paths to the same destination do not create loops

31 Fibbing enables high flexibility Theorem Any set of forwarding DAGs can be enforced by Fibbing fine-grained traffic steering to middleboxes per-destination load balancing for traffic engineering backup paths provisioning

32 Is Fibbing expressive? Does Fibbing scale? Yes!

33 Fibbing can scale and quickly react to failures computing augmented topologies of limited size we have an ILP to strategically place fake nodes pre-computing response to failures to quickly repair augmented topologies applying quickly failure responses relying on the effectiveness of IGP for failure reaction

34 Given a physical topology and a set of path requirements, a linear program computes an optimized virtual topology physical topology forwarding DAGs Integer Linear Program augmented topology + Optimizer minimize topology size

35 Few lies can realize multiple shortest-path deviations (preliminary evaluation on Rocketfuel) # of fake topology components nodes edges # IGP shortest path violations

36 Fibbing enables flexible routing à la SDN, today! Reduce controller concerns most of the heavy work is still done by the routers Maintains operators mental model good old protocols running, easier troubleshooting Facilitates SDN deployment unified interface to routers and SDN switches

37 Sweet Little Lies: Fake Topologies for Flexible Routing Tell me lies, tell me sweet little lies Fleetwood Mac Stefano Vissicchio