Network-Coding based Scheduling And Routing Schemes for SerivceOriented Wireless Mesh Networks Jian-Liang Pan Nguyen Thi Mai Phuong Yu-Chen Chang Tai Yang Wu Tai Long Chen 1 Talk Outline Introduction & Background - Wireless Mesh Networks Mesh vs. Ad-Hoc Networks Wireless Network Coding Routing Strategies Link Scheduling & Channel Assignment Performance Evaluation 2 Wireless Mesh Networks WMN - Wireless Mesh Network: Ad-hoc network with a core which has limited mobility Mesh Router: A wireless base station with limited or no mobility Infrastructure of the network Mesh Clients: A wireless node which is fully mobile, may also act as a router in some WMNs Image: Indigo Systems – WMN for Environmental Monitoring 3 Hybrid WMN 4 Overview Node Types Wireless routers Gateways Printers, servers Link Types Intra-mesh wireless links Stationary client access Mobile client access Mobile clients Stationary clients Internet access links 5 Gateways Multiple interfaces (wired & wireless) Mobility Stationary (e.g. rooftop) – most common case Mobile (e.g., airplane, busses/subway) Serve as (multi-hop) “access points” to user nodes GW Relatively few are needed, (can be expensive) 6 Wireless Router At least one wireless interface. Mobility Stationary (e.g. rooftop) Mobile (e.g., airplane, busses/subway). Provide coverage (acts as a mini-cell-tower). Do not originate/terminate data flows Many needed for wide areas, hence, cost can be an issue. 7 Users • Typically one interface. • Mobility – Stationary – Mobile • Connected to the mesh network through wireless routers (or directly to gateways) • The only sources/destinations for data traffic flows in the network. 8 User – Wireless Router Links • Wired – Bus (PCI, PCMCIA, USB) – Ethernet, Firewire, etc. • Wireless – 802.11x – Bluetooth – Proprietary • Point-to-Point or Point-toMultipoint • If properly designed is not a bottleneck. • If different from router-torouter links we’ll call them access links 9 Router to Router Links • Wireless – 802.11x – Proprietary • Usually multipoint to multipoint – Sometimes a collection of point to point • Often the bottleneck • If different from router-touser links we’ll call them backbone links 10 Gateway to Internet Links • Wired – Ethernet, TV Cable, Power Lines • Wireless – 802.16 – Proprietary • Point to Point or Point-toMultipoint • We’ll call them backhaul links • If properly designed, not the bottleneck 11 How it Works • User-Internet Data Flows – In most applications the main data flows • User-User Data Flows – In most applications a small percentage of data flows 12 Mesh vs. Ad-Hoc Networks Ad-Hoc Networks Multihop Nodes are wireless, possibly mobile May rely on infrastructure Most traffic is user-touser Wireless Mesh Networks Multihop Nodes are wireless, some mobile, some fixed It relies on infrastructure Most traffic is user-togateway 13 Abstract One of critical problem : improve the network throughput - Channel assignment problem Key technology: Network Coding Step1: analyze the throughput improvement obtained by wireless network coding schemes in wireless mesh networks. Step2: develop a heuristic joint link scheduling, channel assignment, routing algorithm 14 Wireless Network Coding 15 Wireless Network Coding Network coding is a particular in – network data processing technique that exploits the characteristics of the wireless medium (in particular, the broadcast communication channel) in order to increase the capacity or the throughput of the network (Network Information Flow – Rudolf Ahlswede, Ning Cai, Shou – Yen Robert Li, Senior Member, IEEE, and Raymond W.Yeung, Senior Member, IEEE) 16 Wireless Network Coding Let G=(V,E) be a graph (G is the set of nodes of a point – to – point communication, E is the set of edges) with source s and sinks t1,t2, …tL. R=[Rij,(i,j) E], Rij is the capacity of an edge (i,j) E. h is the rate of information source F=[Fij,(i,j) E] is a flow in G from s to tl (l=1,…L) if for all (i,j) E : 0 Fij Rij and for all i V except s and tl: i ':( i ',i )E Fi 'i F j:( i , j )E ij 17 Wireless Network Coding F is a max-flow from s to tl in G if F is a flow from s to tl whose value is greater than equal to any other flow from s to tl Max – Flow Min – Cut theorem: Let G = (V,E) be a graph with source s and sinks t1,…,tL, and the capacity of an edge (i,j) be denoted by Rij. (R,h,G) is admissible if and only if the values of a max – flow from s to tl are greater than or equal to h 18 Wireless Network Coding 19 Wireless Network Coding 20 Wireless Network Coding A total of 9 bits are sent (network coding) If network coding is not allowed, at least one more bit has to be sent => Saving 10% bandwidth 21 Wireless Network Coding Using the scheme in fig.8(b), if 2 bits are sent in each edge, then 4 bits can be multicast to all the sinks If network coding is not allowed: Let B = {b1,…,bk} be the set of bits to be multicast to all the sinks. Let the set of bits sent in the edge (s,i) be Bi, where |Bi| = 2, i = 1, 2, 3 B = Bi Bj for any 1 i j 3 We have: B3 ( B1 B2 ) ( B3 B1 ) ( B3 B2 ) B 22 Wireless Network Coding Therefore: k | B3 ( B1 B2 ) | | B3 | | B1 B2 | | B3 | | B1 | | B2 | | B1 B2 | 6k k 3 or if network coding is not allowed, only 3 bits can be multicast to all sinks (if 2 bits are sent in each edge) => Throughput of network can be increased by one-third using a very simple network code 23 Wireless Network Coding 24 Wireless Network Coding 25 Butterfly Network When two or more signal transmission path of partial overlap ,Encoded to improve the signal transmission rate of the technology can be called 「 common path coding」 Each arrow indicates assumed to send a signal , the signal value is 0 or 1,A point given by the two signals x and y,should be sent to the B point and C point. Let M points x and y send a representative to the similarities and differences of the signal , when the B point to receive this signal and x, can be solved y; Similarly, Cpoint can also be solved x. Graph signal x ⊕ y ,x ⊕ y is called 「binary sum」 ,it not only represents a coded form, is also a mathematical "linear" code. Use network coding,the encoding method has three restrictions: 1. Each encoded packet must be the linear independence. 2. The node after the packet encoded into the nodes need to encode packets with each other linearly independent . 3. Enter the node will be re-encoded packets and transmitted. Oct. 23, 2007 28 Linear Network Coding P1,P2 , ‧ ‧ ‧ ‧ ‧ ‧ Pn the packets for the node S e = {e1, e2, e3 ‧ ‧ ‧ ‧ ‧ ‧ en} is coded coefficients Figure (2), (3) is linear combination of matrix encoding. Linear Network Decoding The coefficient matrix is converted into an antimatrix, the re-encoded packet multiplied by the value of the packet can be restored. Network coding of fault tolerance 反矩陣的算法(一) :高斯消去法 步驟: (步驟1) 將單位矩陣In連接於所給矩陣A之後 (步驟2) 利用高斯消去法將步驟1中的A矩陣化成單位矩陣In 求A-1的教學範例: 求出矩陣A的反矩陣, [A In] 反矩陣的算法(二) 伴隨矩陣(adj A) 餘因子(cofactor) Aij: 伴隨矩陣(adj A)的算法 Routing Strategies Two types : Shortest single-path routing Optimized single-path routing 41 Shortest single-path routing ● In particular, single-path routing can be obtained by Dijkstra’s algorithm. 42 Dijkstra’s Algorithm Initial 43 Dijkstra’s Algorithm 44 Dijkstra’s Algorithm 45 Dijkstra’s Algorithm 46 Dijkstra’s Algorithm 47 Dijkstra’s Algorithm 48 Dijkstra’s Algorithm 49 Dijkstra’s Algorithm Finish 50 Optimized single-path routing Select the path that provides the maximum flow in the multipath routing for each session and following two steps: Solve the LP with multipath routing. For each session, we select the path that achieves the highest flow. Input these obtained optimized single-path routes for each session to the LP and re-solve the LP. 51 Link Scheduling Let Ts (Ts ≥ 1) be this period of time in the unit of time slots. Thus, we can update the channel assignment every Ts time slots. 52 Channel Assignment Note that the problem of optimal channel assignments is NP-hard. Therefore design a heuristic greedy algorithm to obtain the schedule, which approximates the optimal solution derived by the LP. 53 Channel Assignment (Cont.) The NC links combinations with larger unassigned flows have higher priority to select the channels. After going through all the NC traffic and all the links for the unicast traffic, we check if there are still unassigned flows. If so, we move on to the next Ts time slots and repeat the scheduling process until all unassigned flows become zero. 54 PERFORMANCE EVALUATION We solve the linear program to obtain the theoretically optimized network throughputs and the corresponding flows for the non-NC scheme , conventional NC scheme, NC3 scheme, and NC5 scheme. multipath routing can significantly increase the network throughput gain as compared to singlepath routing. NC5 and NC3 can increase network throughput by 24 and18 percent 55 The network throughput gain of wireless NC schemes over over the non-NC scheme in the random graph network with 34 nodes the non-NC scheme in the random graph network with 34 nodes. 56 The impact of routing strategies on network throughput. 57 The performance impact of multichannel and multiradio in the random graph network where there are 20 unicast sessions. 58 CONCLUSIONS the analog NC scheme increases by 24 and 34 percent on average as compared to the conventional NC scheme and non-NC scheme, respectively. We modeled the throughput gains of both the conventional NC and analog NC schemes over the traditional non-NC scheme in multichannel multiradio wireless mesh networks. 59 Thank you! 60