Interference-Aware Topology, Capacity, and Flow Assignment in Wireless Mesh Networks

Transcription

1 Interference-Aware Topology, Capacity, and Flow Assignment in Wireless Mesh Networks Eiman Alotaibi Thanks to: Vishwanath, Massimo, Sayeem, Prof. Mukherjee Department of Computer Science University of California, Davis

2 Motivation Topology, Capacity, and Flow Assignment (TCFA) is a general network-design problem Design a cross-layer solution for TCFA That is: - Dynamically allocate resources - Self organized Design a realistic secondary-interference model 5/14/2010 2

3 Why Topology Control? Wireless links are soft (no physical deployment is required) Network topology is not expected to be fixed for a long time A good topology design should follow the traffic demand and assign links as needed to serve as much users as possible within a short period of time Decreasing number of links on a topology decreases interference Decreasing number of links on a topology increases delay (increases number of hops) With minimum number of links, we can assign different channels to each links to achieve the best performance 5/14/2010 3

4 In Wireless Mesh Network Each radio has a limited capacity - This can be used as a constraint instead of Cost Constraint Gateway Rest of the Internet Gateway Wireless Channel is a shared channel Router Router Interference limits the effective capacity Router /AP End users End users 5/14/2010 4

5 Why Topology CFA in WMN? Fully connected Tree Star No. of links high low low Reliability high low low Interference high low high Power high low high 5/14/2010 5

6 Wireless Constraints Primary Interference Constraint: Self Interference Collision Multicast Signal-to-Interference-and-Noise Ratio (SINR) Constraint: G P i,j i,j N I P o p,q,i,j p,q (p,q) L t 5/14/2010 Ref: Channel, Capacity, and Flow Assignment in Wireless Mesh Networks 6 Presentation, by Vishwanath Ramamurthi

7 Wireless Constraints Secondary Interference Constraint: p1 q1 q1 q2 J Link_I(J) = p1 p p2 q2 5/14/2010 7

8 Secondary-Interference Approaches 1. Combined Interfering-links Constraint (1eq) 2. Separate Interfering-links Constraint (Aeq) 3. Multiple Interfering-links Constraint (Meq) 5/14/2010 8

9 Combined Constraint p1 q1 q1 q2 J Link_I(J) = p1 p p2 q2 Cj + Cp1q1 + Cp1q2 + Cp2q1 + Cp2q2 C 5/14/2010 9

10 Separate Constraint p1 q1 q1 q2 J Link_I(J) = p1 p p2 q2 Cj + Cp1q1 Cj + Cp1q2 Cj + Cp2q1 Cj + Cp2q2 C C C C 4 equations 5/14/

11 Multiple Constraint p1 q1 q1 q2 J Link_I(J) = p1 p p2 q2 S1 = { Lp1q1, Lp2q2 } S2 = { Lp1q2, Lp2q1 } Cj + Cp1q1 + Cp2q2 C Cj + Cp1q2 + Cp2q1 C 2 equations 5/14/

12 Secondary-Interference Approaches Our approach (Meq) provides more capacity compared to (1eq) approach (reduce Interference) Our approach reduces number of equations in the MILP compared to (Aeq) approach (increase processing speed) Meq is a More Realistic approach 5/14/

13 TCFA Formulation Given: - Number of nodes and their locations - Number of interfaces per node - Source-destination traffic demands D s,d Minimize: (# of links) + ε (delay) - With respect to: {C i,j } and {λ i,j } Output: - Optimal Network Topology 5/14/

14 Input Notations NA NG N C Dsd Fj Fjf W Ej,Sk α Hmax k Number of routers Number of Gateways NA + NG Maximum radio capacity Traffic demand of a source-destination pair Number of radio interfaces at node j Channel assigned to the f-th radio at node j Number of channels available Set of non-overlapping interfering links at node j Minimum traffic parameter on any link Maximum allowed number of hops along a single(s, d)flow Maximum allowed congestion on any link 5/14/

15 Variable Notations Bij,m Cij Capacity of Lij over channel m aggregate link capacity of Lij over all channels Srci,sd Up/downstream traffic sourced from node i and issued by (s, d) source-destination flow Snkisd Up/downstream traffic sunk at node i and issued by (s, d) source-destination flow ri λij,sd λij λ γij γ hij,sd Total up/downstream traffic that is sourced or sunk at node i Amount of traffic on Lij and belongs to (s, d) flow Total up/downstream flow on Lij over all (s, d) pairs Total traffic on all links (binary) = 1, when Lij carries traffic Number of links selected to represent the new topology (binary) = 1, when Lij is selected to carry traffic along (s, d) flow 5/14/

16 TCFA Model Demand constraints at routers Source node Destination node Intermediate node 5/14/

17 TCFA Model Total flow at a node Throughput calculation Flow-conservation constraints Routing 5/14/

18 TCFA Model Link-flow constraints Delay constraints 5/14/

19 TCFA Model Capacity constraints Primary-interference constraints Secondary-interference constraints 5/14/

20 TCFA Model Link constraints Topology constraints Hops constraints 5/14/

21 Performance Evaluation: Assumptions Single channel Single radio per node Upstream Traffic (40%) Downstream Traffic (60%) At least ½ of the traffic served (feasible solution α = 0.5) 5/14/

22 Performance Evaluation We study different cases We vary: - Number of hops - Number of gateways - Traffic load (per router) - Two objective functions - The value of α (Min traffic on each link) 5/14/

23 Input Topology (mesh) 1 0 GW /14/

24 Result: Interference approaches Number of Links on Topology Gamma Traffic Load (Mbps) Aeq, γ+λ Meq, γ+λ 1eq, γ+λ Aeq, λ+γ Meq, λ+γ 1eq, λ+γ 5/14/

25 Result: Interference approaches Normalaized Network Delay Normalized Delay Aeq, γ+λ Meq, γ+λ 1eq, γ+λ Aeq, λ+γ Meq, λ+γ 1eq, λ+γ Traffic Load (Mbps) 5/14/

26 Result: Interference approaches Network Throughput Throughput (Mbps) Aeq, γ+λ Meq, γ+λ 1eq, γ+λ Aeq, λ+γ Meq, λ+γ 1eq, λ+γ Traffic Load (Mbps) 5/14/

27 Result: Multi-hop Number of Links on Topology (6 Mbps) Gamma No. of hops Aeq Meq 1eq 5/14/

28 Result: Multi-hop Normalaized Network Delay (6 Mbps) Normalaized Delay Aeq Meq 1eq No. of Hops 5/14/

29 Result: Multi-GW Number of Links on Topology Gamma Traffic Load (Mbps) Aeq, 2-GW Meq, 2-GW 1eq, 2-GW Aeq, 1-GW Meq, 1-GW 1eq, 1-GW 5/14/

30 Result: Multi-GW Normalized Network Delay Normalized Delay Traffic Load (Mbps) Aeq, 2-GW Meq, 2-GW 1eq, 2-GW Aeq, 1-GW Meq, 1-GW 1eq, 1-GW 5/14/

31 Result: The value of (α) Normalized Maximum Throughput Alpha Traffic Load (Mbps) Aeq Meq 1eq 5/14/

32 Result: TCFA Efficiency Aggregate Network Throughput 250 Throughput (Mbps) TCFA Mesh Topology Max-throughput Traffic Load (Mbps) 5/14/

33 Conclusion Design a Dynamic and self organize TCFA solution for WMN Deploy realistically the impact of the interference on the link capacity TCFA dramatically improves the performance of WMN The selection of no. of hops is essential 5/14/

34 Q&A Thank you 5/14/

35 Backup Slides 5/14/

36 Cross-Layer Design Transport Flow Control Network Routing PHY/MAC Resource Allocation Scheduling Tx-Rx Interference CA in wireless network should also take into account Interference Interference depends on - Topology - PHY Layer technology - Antenna Beam pattern Benefits of Cross Layer Design PHY layer limitations are considered Network resources are utilized to the best possible extent 5/14/2010 Ref: Channel, Capacity, and Flow Assignment in Wireless Mesh Networks 36 Presentation, by Vishwanath Ramamurthi

37 Complexity Network Design Problems Problem Given Minimize w.r.t s.t CA τ, λ i,j T C i,j D FA τ, C i,j T λ i,j 0 λ i,j µc i,j CFA τ T C i,j, λ i,j D TCFA - T τ, C i,j, λ i,j D - τ = Network Topology - λ i,j = flow on link (i,j) - µ = average packet size - C i,j = capacity of link (i,j) - T = Average System Delay - D = Maximum cost (i,j) E i,j i,j d C D 5/14/2010 Ref: Channel, Capacity, and Flow Assignment in Wireless Mesh Networks 37 Presentation, by Vishwanath Ramamurthi

38 Channel, Capacity, and Flow Assignment (CCFA) Given: - Network Topology, source-destination demands γ s,d - Number of non-overlapping channels K - Number of Network Interface Cards (NICs) on each node q i Minimize: T With respect to: {C i,j }, {λ i,j }, and H i,j {1,,K} 5/14/2010 Ref: Channel, Capacity, and Flow Assignment in Wireless Mesh Networks 38 Presentation, by Vishwanath Ramamurthi

39 Network Utility Efficiency of a WMN Total Throughput Total Demand D D s,d s,d Utility U is defined to include both throughput and delay U T Em Em = Throughput emphasis factor - How much is throughput emphasized over delay Generalized version of Kleinrock s Power of a network 5/14/2010 Ref: Channel, Capacity, and Flow Assignment in Wireless Mesh Networks 39 Presentation, by Vishwanath Ramamurthi

40 Overall CCFA Algorithm Traffic Profiling Channel Assignment Link Flows Channels Multichannel Capacity and Flow Assignment (MCFA) Capacities and Flows U (n) -U (n-1) ε No Yes Channels, Capacities, and Flows 5/14/2010 Ref: Channel, Capacity, and Flow Assignment in Wireless Mesh Networks 40 Presentation, by Vishwanath Ramamurthi