1. Layered Backpressure Scheduling for Delay Reduction in Ad Hoc Networks
Leandros Maglaras
Dimitrios Katsaros (presentation)
Dept. of Computer & Communications Engineering
University of Thessaly, Volos, Greece
WoWMoM Symposium, Lucca (Italy), 21-23/June/2011

2. Wireless Ad Hoc Networks
Wireless Ad Hoc Networks features
Dispersed network
Self-organized
All nodes acts as routers
No wired infrastructure
Potential multihop routes

3. Modern battlefield: Delay does matter

4. Packet scheduling based on Back-pressure: the LayBP protocol
Our proposal …
Packet scheduling based on Back-pressure: the LayBP protocol
Creates clusters based on dense connectivity
Forwards packets on a cluster-by-cluster basis, until the packet enters the ‘destination’ cluster
Reap the benefits of both worlds:
Cluster-based backpressure
Shortest-path backpressure

5. The back-pressure protocol
Each node:
Maintains one queue for every other node c a e d
= Optimal Commodity for link (e,d) on slot t (maximizes diff. backlog)

6. Backpressure: pros-and-cons
throughput-optimal routing algorithm
robust to topology changes
Based on solid mathematical background (Lyapunov drift theory), thus, ability to study stability properties
Backpressure ↓:
Its performance deteriorates in conditions of low, and even of moderate, load in the network, since the packets “circulate” in the network:
the backpressure stabilizes the system using all possible paths throughout the network
The net effect of this mechanism is to increase delay (and thus energy consumption)

7. Delay-aware backpressure: Cluster-based
@ INFOCOM08
Motivation: Reduce the number of queues
How?
Define arbitrary clusters
Intercluster communication via gateways
Maintain one queue per gateway at each relay node which still leads to an excessive number of gateways
Delay increases with increasing clusters number!
Why? Too many gateways

8. Delay-aware backpressure: Shortest-path
@ INFOCOM09
& IEEE Trans. on Net 11
Motivation: Delay can be reduced via shortest-path routing
How?
Precomputation of all pairwise-node distances
application of the backpressure notion on the shortest path(s) between source and destination
Achieves the lower-bound of delay
Removal of a single link keeps delay high!
Why? Shortest-path recalculation is needed

9. Can we combine the two?
‘Relax’ the demand for precomputation of shortest-paths
Robustness to topology changes
Maintain the idea of ‘pushing the packets to the preferred direction’
Exploit the clustering idea, but
Scale-up to the number of clusters and/or gateways
Smart clustering

10. Our proposal: Layered Backpressure
LayBP ideas:
[1st] Split the network into layers according to the connectivity among them
usually implies geographic proximity, as well
[2nd] Assign layer ID respecting geo-proximity
Breadth-first or fractal curve (Hilbert) numbering
[3rd] Forward packets according to the destination layer ID
The layers play the role of attractors, that attract the packets destined for them and then “disallow” the packets to leave the layer

11. Layered Backpressure: Packet scheduling
Each node n maintains a separate queue of packets for each destination
The length of such queue is denoted as Qdn[t]
For every queue Qdn[t] the node computes the parameter Dleveldn which represents the absolute difference between current and destination node’s layer number
Dleveldn = |Layer(n) − Layer(d)|

12. LayBP operation

13. Throughput optimality
Following the work:
M. J. Neely, E. Modiano, and C. E. Rohrs. Dynamic power allocation and routing for time varying wireless networks. IEEE JSAC, 23(1):89–103, 2005
Since, parameter Anmd is bounded (for LayBP it is a constant), the policy is throughput optimal

14. Evaluation setting LayBP compared to:
Backpressure (BP), Shadow-Queue backpressure (SQ-BP), Cluster-based backpressure (CB-BP), Shortest-path backpressure (SP-BP)
Performance measures: Delay & Throughput
Network topologies:

15. Parameter tuning
A=2 is OK

16. Few gateways: More rapid convergence
Delay performance
Many gateways
grid
Few gateways
Few gateways: More rapid convergence

17. Throughput performance
All policies are throughput-optimal

18. Summary Wireless Ad Hoc Networks New Packet Scheduling protocol, LayBP
LayBP evaluation and comparison to the state-of-the-art protocols
LayBP:
throughput-optimal
tradeoff between delay and calculations
robust to topology changes

19. Relevant references
L. Tassiulas and A. Ephremides. Stability properties of constrained queueing systems and scheduling policies for maximum throughput in multihop radio networks. IEEE Trans. Automatic Control, 37(12):1936–1948, 1992.
L. Ying, R. Srikant, and D. F. Towsley. Cluster-based backpressure routing algorithm. In Proc. INFOCOM, pages 484–492, 2008.
L. Ying, S. Shakkotai, and A. Reddy. On combining shortest-path and back-pressure routing over multihop wireless networks. In Proc. INFOCOM, And IEEE/ACM Trans. On Networking, 19(3): 841–854, 2011.

20. Thank you for your attention!
Questions?

21. Topology change
Inactive one of the two intercommunity links

22. Impact of topology change

23. Better layer naming: Hilbert curves

24. For 3-D wireless sensornets: Underwater

25. Clustering: max-min d-cluster formation
@ INFOCOM04 &

26. Original backpressure