mon mission optimized overlay networks
play

MON: MISSION-OPTIMIZED OVERLAY NETWORKS Bruce Spang , Anirudh - PowerPoint PPT Presentation

MON: MISSION-OPTIMIZED OVERLAY NETWORKS Bruce Spang , Anirudh Sabnis, Ramesh Sitaraman, Don Towsley, Brian Decleene SETTING Managed by a single organization Good model of underlying network Di ff erent types of tra ffi c use this


  1. MON: MISSION-OPTIMIZED OVERLAY NETWORKS Bruce Spang , Anirudh Sabnis, Ramesh Sitaraman, Don Towsley, Brian Decleene

  2. SETTING ➤ Managed by a single organization ➤ Good model of underlying network ➤ Di ff erent types of tra ffi c use this network ➤ The organization cares more about certain types of tra ffi c

  3. MON

  4. FLOWS ➤ A flow corresponds to tra ffi c being routed between unique pair of source-destination, tra ffi c type (ex. VOIP , Video). ➤ A flow could use multiple overlay paths. ➤ A flow could have multiple connections. ➤ Example . 10 VOIP connections from BAE to UMass

  5. UTILITY FUNCTIONS ➤ Utility Function is a measure of “usefulness” for a certain rate ( x Mbps) allocated for a single connection of the flow. ➤ Increasing function, possibly non-convex

  6. MON OPTIMIZATION PROBLEM ➤ U k (x) - Utility of flow k ➤ k - a flow ➤ n k - number of connections ➤ C l - estimated capacity ➤ x k,r - rate of flow k on path r ➤ N k - estimated demand

  7. PFO Solve the problem o ffl ine using Global Optimization Techniques

  8. MOTIVATION FOR CONGESTION CONTROL ➤ Solving optimization problem may take time ➤ What if network changes? Or demand changes? ➤ Quickly react to packet loss and increased bandwidth ➤ We use TCP for this

  9. MONTRA OPTIMIZATION MODEL

  10. TCP - INCREASE/DECREASE RULES ➤ x f - sending rate of flow f ➤ γ - stability constant ➤ w f - weight

  11. HOW TO PICK W F ? ➤ Controllers should match PFO’s target rates ➤ Controllers should make good decisions if network changes

  12. MAPPING PFO TO MONTRA Theorem Idea: Assume PFO has an accurate model of the network. Suppose PFO picks a target rate A. If the gradient of MONtra matches the gradient of PFO at A, then MONtra will converge to A Proof: See the paper

  13. MON

  14. UTILITY FUNCTIONS AND FLOWS ➤ Used following utility functions: Flow A Flow B ➤ Generated random flows

  15. DOES MONTRA SEND AT THE CORRECT RATE? Site r15 Router r16 s30 r17 r19 r18 s31 r20 s33 s32 r0 r21 s34 r2 r1 r4 r22 r3 s25 r5 r23 r10 r8 r6 s26 s35 s27 r24 r11 s29 r9 r7 s28 r12 r14 r13 MONtra converges to PFO’s rates on a complex network

  16. BENEFITS OF OVERLAY ROUTING Number of hops : Mission utility increases with number of hops Number of paths: Allowing more paths increases utility

  17. HOW DOES MONTRA REACT TO SLIGHT CHANGES IN NETWORK? x Mbps 10 Mbps 5 Mbps • Ran PFO with x=3 Mbps • Adjusted capacity from 1 to 9 Mbps. • MONtra does better than baselines

  18. HOW DOES MONTRA REACT TO NETWORK FAILURES? PFO recovers Triangle topology: Link fails x

  19. HOW DOES MONTRA REACT TO NETWORK FAILURES? AT&T topology: • Link failure decreases utility • PFO recovers utility

  20. SUMMARY ➤ Overlay network architecture ➤ Optimally route flows ➤ Resilient to changes to inputs

  21. THANKS!

  22. QUESTIONS?

Recommend


More recommend