Presentation is loading. Please wait.

Presentation is loading. Please wait.

An Adaptive, High Performance MAC for Long-Distance Multihop Wireless Networks Sergiu Nedevschi *, Rabin K. Patra *, Sonesh Surana *, Sylvia Ratnasamy.

Similar presentations


Presentation on theme: "An Adaptive, High Performance MAC for Long-Distance Multihop Wireless Networks Sergiu Nedevschi *, Rabin K. Patra *, Sonesh Surana *, Sylvia Ratnasamy."— Presentation transcript:

1 An Adaptive, High Performance MAC for Long-Distance Multihop Wireless Networks Sergiu Nedevschi *, Rabin K. Patra *, Sonesh Surana *, Sylvia Ratnasamy #, Lakshminarayanan Subramanian † and Eric Brewer * * UC Berkeley EECS Department # Intel Research Berkeley † NYU, CS Department ACM MobiCom 2008

2 Outline Introduction Protocol Design Evaluation Conclusion

3 Introduction Multi-hop WiFi long-distance networks (WiLD) provide cost efficient connectivity to  Sparsely populated areas  Rural regions in developing  Industrialized countries

4 Introduction Due to the presence of long-distance links and their use of directional antennas, WiLD networks present unique challenges:  Multihop propagation delays  Increased likelihood of inter-packet collisions Prior work has shown that these challenges make traditional MACs based on carrier sensing, a poor fit for WiLD networks.

5 Related Work B. Raman and K. Chebrolu. “Design and Evaluation of a new MAC Protocol for Long-Distance 802.11 Mesh Networks.” In ACM MOBICOM, Aug. 2005.

6 Using SynOp to avoid Interference a)Mix-Rx-Tx R1 receives and T2 transmits Not feasible because the signals that T2 and T1 transmit interfere with each other at receiver R1tim b)Syn-Rx R1 and R2 receive simultaneously Feasible physically c)Syn-Tx T1 and T2 transmit simultaneously Feasible physically 3 possible operating ways of node A. Node A has two transceivers for links with B and C, respectively.

7

8 Related Work (MAC protocols for WiLD links)

9 Related Work AC BD T=20 T=14 T=6 t=20 t=14 t=6 t=20 Total Transmission Time slot = 20+14

10 Allowing neighboring transmissions that overlap AC BD T=20 T=14 T=6 t=6 Total Transmission Time slot = 20 t=20

11 Motivation and Goal Motivation  These solutions rely on a TDMA schedule with fixed- length slots and hence cannot adapt to dynamic traffic variations. Goal  To design a dynamic slot adaptation enables more efficient use of network capacity by adapting to traffic.

12 JazzyMac Protocol Design Token (T ij ): the node holding the token can transmit on the associated link. AB T AB Timeout value (v ij ): the node holding the token is allowed to transmit over the associated link T AC C T AB v AB =10

13 Four rules of JazzyMac Token exchange rule Mode rule:  A node i that is in receive mode can transition to transmit mode only when it holds the token for all its links. Transmission rule:  A node i is in transmit mode.  A node i holds token T ij, and T ij is valid. Slot rule:  A node i can transmit on link ij for no longer than max_slot time units.

14 Protocol Bootstrapping Algorithm  Color the vertices of the network graph with the minimum number of colors K such that no two adjacent vertices have the same color.  The tokens are assigned to the link end that has the lowest color/weight (the two ends must be colored differently ). A BC : 1 : 2 : 3 T AB T AC T BC weight

15 A BC T AB T AC T BC T AB T AC T BC v AB =35 v BC =15 v AC =15

16 JazzyMac Properties S i (A): the sequence number of node A T i : the set of nodes transmitting during slot i, T i ⊆ G The condition to be fulfilled by node A in order for A to belong to the set T i can be expressed as:

17 JazzyMac Properties (cont.) Property 1. During any time slot, the difference in sequence numbers between any two network nodes remains strictly smaller than the number of colors used for graph coloring: Proof

18 JazzyMac Properties (cont.) PROPERTY 2.  There protocol does not result in any deadlock or node starvation. PROPERTY 3.  Every node can choose to send on each of its links for at least 1/K of the link capacity. PROPERTY 4.  The maximum delay between two consecutive opportunities to send on any link is smaller than 1/K

19 Dealing with Loss AB T AB S AB = 4S AB = 5 K × max_slot T AB S AB = 4 T AB S AB = 4

20 Evaluation FT: Fixed-slot TDMA according to vertex colors.  First compute the minimum vertex coloring of the graph.  Then nodes transmit in TDMA slots, according to their color.  Colors are scheduled for transmission in a round-robin fashion. FT-CUT: Fixed-slot TDMA over maxcut.  First compute the maximal subgraph that is bipartite and contains all the network nodes (i.e. a maxcut in the original graph.)  Then use 2P on the maxcut, keeping other links as backups.

21 Setting SimulatorJava-based simulator Topologies Random topologiesvarying degrees of connectivity and sizes An actual real-world topologyAavind Eye Hospital in India Typical mesh WiFi topologiesRaman topology Traffic flows sizeCBR, 500Kbps Link capacity10Mbps Slot size20ms patterns of traffic demand one source to many randomly distributed destinations unidirectional CBR flows, with randomly chosen source destination pairs A pairs of CBR flows in opposite directions (bidirectional)

22 Number of good flows as we add unidirectional CBR flows

23 Average delay of good flows as we add unidirectional CBR flows

24 Throughput for various topologies, random CBR flows from one source to all the nodes

25 Throughput with Unidirectional random CBR flows

26 Throughput with Bidirectional random CBR flows

27 Maximum throughput vs. average delay

28 Conclusion This paper presents JazzyMac, a fully distributed, practical MAC layer that uses local traffic information to adapt link transmission slot sizes dynamically. It exploits asymmetric traffic, time varying traffic, and non-bipartite topologies to achieve a much higher throughput than existing TDMA-based approaches.


Download ppt "An Adaptive, High Performance MAC for Long-Distance Multihop Wireless Networks Sergiu Nedevschi *, Rabin K. Patra *, Sonesh Surana *, Sylvia Ratnasamy."

Similar presentations


Ads by Google