1. Rate-Based Channel Assignment Algorithm for Multi-Channel Multi- Rate Wireless Mesh Networks Sok-Hyong Kim and Young-Joo Suh Department of Computer Science & Engineering Pohang University of Science & Technology (POSTECH), Pohang, Korea IEEE GLOBECOM 2008

2. Outline  Introduction  Motivation  Rate-Based Channel Assignment algorithm Spanning Tree Formation Rate-Hop and Local Traffic Count Metrics for RB-CA Channel Assignment Algorithm  Performance evaluation  Conclusion

3. Introduction  A WMN consists of a set of stationary wireless mesh routers forming a multi-hop wireless backbone  Some routers act as gateways linked to the Internet via high- speed wires  Each mesh router offers Internet access for wireless mesh clients

4. Introduction  The network capacity and the performance of WMN can be increased by using multiple non-overlapping channels  As a low-rate link tends to occupy the wireless channel longer than the high-rate link, the overall performance of network is subject to low-rate links  This problem is often referred to as performance anomaly

5. Motivation - performance anomaly All routers in the figure form a spanning tree, where node A is the parent of nodes B, C, and D We assume that the total number of channels is three The offered load is enough on all the links

6. Motivation - performance anomaly Each flow starts at t = 10, 15, and 20 seconds from nodes B, C, and D, respectively

7. Rate-Based Channel Assignment algorithm  Our focus is assigning a channel and selecting a path based on the data rate of possible links in WMNs  In general, each router joins one or multiple gateways by constructing one or multiple spanning trees. However, we restrict routers to join only one spanning tree  And they are equipped with two interfaces: parent interface and child interface

8. Rate-Based Channel Assignment algorithm  Spanning Tree Formation  Rate-Hop and Local Traffic Count Metrics for RB-CA  Channel Assignment Algorithm

9. Spanning Tree Formation  In WMNs, each router on the tree periodically broadcasts Mesh Advertisement (MA) messages to its 1-hop neighbors on every channel  Initially, the gateway connected to the Internet via wires only broadcasts MAs. After a while, nodes that joined the tree can send periodic MAs  A MA includes the source address of the MA, internet accessibility, channel number of the child interface, and the primary and secondary metrics.

10. Spanning Tree Formation (1) A X Y B gateway X sends periodic Mesh Advertisement (MA) to its 1-hop neighbors Y joins node X by sending a Mesh Join (MJ) message which includes its own address X creates an entry for child node Y and replies with Mesh Acknowlegment (MAck) X forwards the received MJ message to the corresponding parents on the path to the gateway. Then, each parent adds Y ’ s entry to its routing table

11. Spanning Tree Formation (2) A X Y B gateway X sends periodic Mesh Advertisement (MA) to its 1-hop neighbors Y joins node X by sending a Mesh Join (MJ) message which includes its own address X creates an entry for child node Y and replies with Mesh Acknowlegment (MAck) X forwards the received MJ message to the corresponding parents on the path to the gateway. Then, each parent adds Y ’ s entry to its routing table Y sends a Mesh Disjoin (MD) message to its previous parent (say node B) after it receives MAck Accordingly, all parents of node B delete the entry for child node Y from their routing tables

12. Rate-Hop and Local Traffic Count Metrics for RB-CA  We consider a MCMR WMN based on IEEE 802.11b. The available data rates are 11Mbps, 5.5Mbps, and 2 Mbps.  The primary metric is rate-hop (RH), which reflects the data rate of a path on the tree and a smaller RH is preferred  The secondary metric is Local Traffic Count (LTC)

13. rate-hop (RH) Since a higher data rate provides better performance, three RHs have the following relationship:

14. rate-hop (RH) Note that low-rate links significantly degrade network performance due to the performance anomaly. Thus, the priority :

15. rate-hop (RH) But, in terms of RH, smaller RH has higher priority. Thus, we have:

16. rate-hop (RH) In practice, for specific values for RH 11Mbps, RH 5.5Mbps, and RH 2Mbps, we set RH 11Mbps to a positive constant. Then, RH 5.5Mbps and RH 2Mbps in eq. (2) have the lower bound. Thus, we restrict the range of RH 2Mbps as follows: where k is a tunable parameter to set the upper bound on RH 5.5Mbps and RH2Mbps in eq. (2). Moreover, a path which consists of more than five 11Mbps- links supports lower data rate than 2 Mbps-link. Thus, k is set to five

17. rate-hop (RH) We can obtain the RH metric of node n expressed as Each node maintains the RH in the routing table entry for data forwarding from or to the gateway update the metric with the recent RH in a MA from its parent. The updated RH is advertised by nodes on the tree via periodic MAs. RH 11Mbps, RH 5.5Mbps, and RH 2Mbps are assigned to 10, 21, and 43, respectively, based on eqs. (1)~(3).

18. Local Traffic Count (LTC) In a case that a node discovers multiple parents having equivalent RH metrics A node n on the tree locally computes LTC by adopting the following weighted averaging technique, where α ∈ [0,1] Every node periodically counts the amount of transmitted and received traffic in bytes through its child interface Then, it updates the amount as LTC new and broadcasts the updated LTC metric via periodic MAs.

19. Local Traffic Count (LTC) Since traffic loads typically converge on the gateway in WMNs, the traffic variation also rapids near the gateway whereas that of nodes in the last hop is relatively slow. Thus, the nodes near the gateway use a large α value to reflect the up-to-date traffic loads as much as possible.

20. Channel Assignment Algorithm  Initially, node X assigns random channels for its parent interface and child interface  Upon receiving the MAs from several nodes on the tree, node X finds the “ best parent node ” based on the RH metric in the MA to assign a proper channel for the parent interface  Node X finishes channel assignment for its parent interface, it then selects the least loaded channel for its child interface to reflect the current channel condition.

21. Channel Assignment Algorithm  In a distributed manner, physically r-hop neighboring nodes on the tree periodically exchange Channel Traffic Count (CTC) messages and maintain CHannel Table (CHT) which contains the load information per channel  Then, each node periodically checks the least loaded channel and changes the current channel of its child interface if necessary  When a node decides to change its child interface ’ s channel, it informs child nodes of the new channel via CHannel Change (CHC).

22. Performance evaluation  ns-2  network size is 2500 m by 2500 m  a single gateway and several (5 to 30) routers  All nodes are randomly distributed and each node is equipped with two interfaces  The transmission and interference ranges are 250 m and 550 m  The number of available channels is 12  Each node uses 11, 5.5, and 2Mbps based on the 802.11b corresponding to the ranges of 100, 200, 250 m, respectively

23. Performance evaluation

25. Conclusion  We proposed the Rate-Based Channel Assignment (RB-CA) algorithm for MCMR WMN.  The proposed algorithm alters a low-rate single-hop path to a high-rate multi-hop path.  the proposed scheme shows improved performance in packet delivery ratio and end-to-end delay

26. Thank You !