Buffered Based Routing and Resiliency Approach for WMN

Transcription

1 Buffered Based Routing and Resiliency Approach for WMN 1 eetanjali Rathee, Ankit Mundra (MIEEE) 1, Department of Computer cience and Engineering Jaypee University of Information Technology Waknaghat, India eetanjali.rathee1@gmail.com, ankit.mundra@ieee.org Nitin Rakesh (LMCI, MIEEE), 4.P. hrera (FBC, MIEEE), 4 Department of Computer cience and Engineering & ICT Jaypee University of Information Technology Waknaghat, India nitin.rakesh@ieee.org, 4 sp.ghrera@rediffmail.com Abstract everal approaches to effectively perform communication in Wireless Mesh Network (WMN) have been proposed. Further communication failures and possibilities of minimum services during such failures are very hard to maintain. Recently resilient multicasting and Resilient Opportunistic Mesh Routing for WMN (ROMER) approaches proposed routing and resilient methods over WMN. Resilient multicasting approach establishes two node disjoint path between each source-destination pair. But this approach can immune from any single link or intermediate node failure. While ROMER approach provides redundant copies of data in randomized manner to increase WMN throughput. We have proposed Buffer Based Routing method in which instead of maintaining routing table at each node we provide buffering at each node which reduces routing cost. At specific interval of time (after specified communication steps) it clears the buffer to increase performance and prevent overhead in network. Using this approach we effectively perform resiliency and exploits advantages of WMN. The performance of the proposed approach is studied by evaluating the cost involved in routing and resiliency using these approaches and compared with the proposed approach. Index Terms WMN; Resiliency; Buffers; Routing I. INTRODUCTION When considering a Wireless Mesh Network (WMN) [1], resiliency [] is becoming a key requirement and service primitive for various applications. everal algorithms [-11] based on resiliency for WMN have been proposed. We have considered two of these recently proposed approaches; 1) Resilient Multicasting in WMN [1]; ) Resilient Opportunistic Mesh Routing for Wireless Mesh Networks (ROMER) [1]. To examine resilient multicasting approach [1] over WMN, we consider a WMN (figure 1) with source node and multiple destination nodes D = {,,... } to reach from source to destination. Resilient multicast [1] will select at least two disjoint-node path for each {, D} pair. In case of failure,traffic will switch to another path to reach the destination. But one of the major drawbacks of resilient multicast is that it increases the network congestion and can immune from any single node/link failure in network. While ROMER [1] define credit based forwarding approach to transmit data from source to destination D. ource node will forward the packet by taking maximum credit cost. At each node, value of and will be calculated; where is the credit cost to reach from one node to its sink node and is the threshold value. ource node will forward data with maximum credit to its sink node. As the sink node will get the data it will calculate value of and compare value of and. If then only node will forward the packet, otherwise it will simply discard the packet. The drawback of this algorithm is that if node cost is higher, there is a possibility to discard the packet even at initial stage. In this paper we studied various drawbacks of these approaches and proposed a Buffer Based Routing Approach which removes disadvantage of previous approaches and is more advantageous than these two approaches. Fig. 1. Example of a Wireless Mesh Network. The paper is divided in five sections. In section two we demonstrate execution of recently proposed approaches: Resilient Multicasting in WMN [1] and ROMER [1] and study their performance using a common network example.. In section three we have proposed our Buffer Based Routing (BBR) approach. In section four we analyze the performance of proposed approach. Finally we conclude the paper and discuss the future scope of the work. II. COMPARATIVE ANALYI OF PREVIOU APPROACHE In this section we have compared the performance of previous approaches (Resilient Multicasting in WMN [1] and ROMER [1]). This comparative study evaluates disadvantage of these popular approaches to communicate data in WMN. For this study we have considered a WMN as shown in figure 1. The B F

2 network consists of one source () node and three destination,, nodes. Remaining nodes of this network are intermediate nodes. We have firstly evaluated [1] and then [1]. Resilient multicast [1] proposed resiliency based routing approach with two node disjoint-paths for each, pairs. Let us examine this approach using WMN as in figure1 where,, are three disjoint-paths between,. The possible steps to communicate data between, in WMN using [1] are path estimation and failure recovery. tep1: Path Estimation This step calculates two node disjoint-paths for each,. Output: possible disjoint-paths for figure 1 are as in Table 1: Path\, TABLE I. PATH EVALUATION ource Destination Pair 1 According to Table1 from there exist two paths and [ and packets will be flooded over both paths concurrently. This approach will increase traffic over two node disjoint path to send packet to a destination. The major drawback of this approach is increase in network congestion. Another drawback with this approach is that it estimates only first two disjoint paths even if other shortest paths exist. tep : Failure Recovery In previous step two distinct paths are generated to communicate from source to all three destinations,, (as shown in Table1). Failures may always exist in network during communication. Resilient multicast considers two types of failures in network i.e. node failure and link failure. To explain node/link failure and its recovery let us consider an example where source is and destination is. Now according to step1 there exist two disjoint paths and. B F B F The below given cases show the consequences of node/link failure and its recovery using resilient multicast approach [1]. Case 1: (Node Failure) initially both and paths were followed concurrently to communicate the data between. If node fails, destination node switches to the unaffected path i.e..imilarly will follow and will follow path. Thus even during failure of node, data communication from source to destination,, can be achieved (see figure ). Case : (Node Failure) if node C fails, in of case, traffic will switch to the path that excludes node C i.e. it will follow path to reach. imilarly } will follow and will follow path. Therefore during node failure the approach will switch to another path (as evaluated in step1) which excludes the failure node to communicate between source to destination (figure ). Case : (Link Failure) if link between fails, then it will affect, } and. Traffic will automatically switch to unaffected path i.e. to reach. is the path for while is the path for. Case 4: (Link Failure) if link fails, then path will switch to, } will follow and will follow. If both links and fails then there is no path available to communicate between,, and network communication will fail. imilarly if nodes and fails (in case 1 and ) then entire communication fails. Furthermore, there will be no possible path available to reach to destination except nodes and C which means it can immune at most one node/link failure in the network. Now let us evaluate ROMER [1] which is another routing approach for WMN. It describes credit based forwarding approach to reach from,.where is the source node and M, M and M 5 are intermediate nodes and D is node (see figure ) M M M 5 Fig.. Resilient Opportunistic Mesh Routing for Wireless Mesh Network (ROMER). B F B F Fig.. Failure Cases in Wireless Mesh Network. Each node is assigned some cost and the packet traverses from one node to another by consuming this cost. To explain this approach let us assume that cost of source node is 100 units and credit limit is also 100 units. Now packet may consume 00 units of cost to reach from.the

3 possible steps to communicate data between, in WMN using [1] are as follows: tep 1: Let us assume that to reach from node to M (intermediate node), packet has consumed 10.5 units. Node M broadcasts packet by consuming unit of cost to its next node i.e. M and M 5. At each node the value of remaining credit ratio ( ) and threshold ( ) are computed and compared. Let us discuss cases of computation at node M which means that there exist two paths i.e. M M and M M 5. We have shown it using two cases: Case 1: (when ) i.e. discard the packet.the cost of M is 50 units, M has 55 units. In this case it will follow M, M path. o, M will forward packet consuming unit of cost to M. After receiving the packet, M will calculate R and T. ee detail as below. At M,. / (1) 77. / () 77. /. /. Where ( ) =77.5 is the amount of credit needed; ( 100-(77.5)) =.5 is the remaining credit available for M and.5/100 is the ratio of remaining credit to initial credit. Further threshold value is calculated as T = =.05, where 55 is M cost and 100 is the cost of the source. Case : (when R>=T) i.e. packet will be forwarded to next node. The cost of M is 50 units, M 5 has 50 units. In this case it will follow M M 5 path. o, M will forward the packet by consuming unit of cost to M 5. After receiving the packet, M 5 will calculate R and T. At M 5 R= (100-( ))/100 () 7. / 7. / 7. /. 7 Where. 7. is the amount of credit needed.. is the cost consumed by packet by reaching from source node to intermediate node (M ) ;50 is the cost of node M 5 and is the source cost (1)) ; is the remaining credit available for M 5 and 7. /. 7 is the ratio of remaining credit to initial credit. The threshold value T= =.500, where 50 is M 5 cost and 100 is the cost of the source. Now, here, so packet will be forwarded to next node. The major drawback of ROMER[1] approach are;1) discarding the packet (when the node cost becomes higher, see figure ); )during node or link failure the possibility of successful packet delivery till the destination node reduces based on the predefined credit cost; comparative study of resilient multicast [1] and ROMER [1] in WMN is shown in table. In section we have proposed Buffer Based Routing (BBR) as a solution to overcome these disadvantages. TABLE II. COMPARION BETWEEN REILIENT MULTICAT AND ROMER.No. Resilient Multicasting in WMN ROMER 1. It follows two node disjoint-path It follows credit based approach, in which packet forwards concurrently on both paths to reach the destination.. Resilient multicast can tolerate maximum of 1 node/link failure in network.. Network congestion problem occurs as it broadcasts data concurrently on both paths. forwarding approach in which packets reached to destination by consuming credit cost. During node or link failure the possibility of successful packet delivery till the destination node reduces based on the predefined credit cost. Discarding the packet when the node cost becomes higher. III. BUFFER BAED ROUTIN APPROACH In previous section we have evaluated several drawbacks allied with both resilient multicasting [1] and ROMER [1] approaches in WMN. In order to provide an efficient solution over these drawbacks, we have proposed Buffer Based Routing (BBR) approach which adopts routing technique based on buffer allocation. The routing approach starts with the selection of the route with minimum number of buffered nodes. The proposed approach consists of three steps: i) buffer allocation to the network nodes; ii) selecting optimum path for routing; and iii) resilient packet transmission. A. Buffer Based Allocation Buffers are placed alternatively according to least cost path selection. The buffer allocation is achieved in five steps: 1. elect a node randomly, where =1,, in network, assign buffer to it and mark it as visited node.. Choose least cost path of node and move to next node, make next node as, skip buffer assignment to that node and mark it as visited node.. Again choose least cost path and move to next node, make next node as, assign buffer at that time.(place buffer at alternate positions) and mark it as visited node. 4. If next node are _ is visited then rollback to its parent node else move forward. 5. Repeat steps until total buffer placement in network are where are number of nodes in the network. B. Buffer Based Routing In BBR approach we have customized the routing table according to include the buffer field with respect to node. The routing table consist of seven parts i.e. 1) node; ) node address; ) next hop; 4) next hop buffered?; 5) next hop address; 6) cost; and 7) buffered node?. The construction of routing table involves following steps: 1. Columns node id; node address; next hop id; next hop address; and cost; describe the id of current node, address of the node, id of next hop, address of next hop, cost to the next hop respectively. We do not go in detail of the explanation for updating these fields.. Now, the remaining two fields i.e. next hop buffered?; buffered node?; are initialized with the value No.

4 These fields are updating recurrently with the buffer allocation process (shown in Table and Table 4).. When the buffer allocation algorithm terminates, the last node broadcast the information of these two fields in the network. 4. All the nodes update their routing table accordingly. We have shown the routing table of two nodes i.e. source node and destination node d. We assume that we start the buffer allocation process form node. The routing tables of each node are updated as the buffer allocation process progresses. For example table shows the routing table of node. Node is the starting node of buffering process so the entries of column Next hop buffered and Buffered node is updated accordingly. Table 4 shows the routing table of node. It is the last buffered node, hence it contain the latest updated routing table. Further, after termination of algorithm node broadcast the routing table information (of two columns i.e. Next hop buffered? and Buffered node?) to entire network. After receiving the updated information each node updates its routing table according to table 4. TABLE III. NODE A ROUTIN TABLE Node Node Next Next hop Next hop Cost Buffered ID address hop ID buffered? address node? A C No 1 Yes A D No Yes B D No 1 No B No No B E No 1 No C F No/Yes No C H No No D A No/Yes No D B No/Yes 1 No D F No/Yes No D No 1 No E B No/Yes 1 No E J No/Yes No F C No No F D No No F H No 4 No F I No No D No 1 No B No/Yes No I No 1 No J No/Yes 1 No H C No No H F No/Yes 4 No I F No/Yes No I No 1 No J No 1 No J E No No imilarly the routing table of each node have constructed. Here we shown the routing table of an intermediate node i.e. D. TABLE IV. NODE J ROUTIN TABLE Node ID Node address Next hop ID Next hop buffered? Next hop address Cost Buffered node? A C No 1 Yes A D No Yes B D No 1 Yes B No Yes B E No 1 Yes C F Yes No C H No No D A Yes No D B Yes 1 No D F Yes No D No 1 No E B Yes 1 No E J Yes No F C No Yes F D No Yes F H No 4 Yes F I No Yes D No 1 No B Yes No I No 1 No J Yes 1 No H C No No H F Yes 4 No I F Yes No I No 1 No J No 1 Yes J E No Yes Figure 4 shows the buffer allocation process in our network. Buffer allocation algorithm tep 1 a) select random node N i from the network. b) Assign buffer (N i ); c) Visited list [] =N i ; tep d) if (next [Ni]! =null) e) Ni min_cost_next [N i ]; f) If (Ni=visited list []) g) elect next_min_cost (N i ); h) o to step (d); i) end if; j) Else k) If (previous [Ni] =buffered array []) l) kip (N i ); m) Visited list [] +1=N i ; n) oto step (d); o) end if; p) Else q) Allocation(N i ); r) otostep(d); s) end else; t) end else; u) end if; v) Else w) Rollback (N i ); x) end else; tep Repeat step until two rollbacks occur at the same node; tep 4 Return; Allocation (Ni) begin if (Ni!=visited_list[]) assign_buffer(n i );

5 visited_list [] +1=N i ; end if; else rollback (previous N i ); end else; end; assign_buffer (Ni) begin assign buffer to N i ; array_buffer [] N i ; update_routing_table(); end; C. Buffer Based Resilient packet transmission In our approach we have customized buffer based routing path along with packet content. To understand resilient packet transmission, let us consider a WMN of 10 nodes (figure 4). Here source and destination are randomly selected throughout the network. Our aim is to send data packets from the source node A to destination node J. For packet transmission the routing path need to have following characteristics: 1. The route must contain minimum number of buffered node.. If more than one path has same number of buffered node than it will select the least cost path. The acknowledgement (ack) message consist ack as well as the id of sender node. Round trip time is defined (RTT) as transmission time of packets from one node to its succeeding node and then receiving acknowledgement back from that succeeding node. Routing form node to node is progressed in following steps: tep 1: According to buffer allocation process for WMN network as in figure 4 is the buffered node so it store packets until it receive acknowledgement from the next buffer node. By considering its routing table, it will select path [ and wait for ack from node and buffered node. tep : Node is the next hop so it receives the packet and lookup the routing table for the next hop which tends towards destination. As it is a non-buffered node it forwards the packet to the next node i.e. and send ack to node. tep : When node receives the packet it stores that packet into its buffer. Further it sends ack to node, and forward the packet to next node i.e.. Meanwhile after receiving the ack from node. Node will send an ack to previous buffered node. Further node deallocate packet from its buffer. tep 4: After receiving the packet, node forwards this to next hop i.e. node. tep 5: Node forwards the packets to node which is the destination node. Note: In our approach the ack is send by both buffered and non- buffered nodes. But buffer node stores the data packet until it receives the ack of next buffered node. It confirms the delivery of message to the next buffered node. Following cases: (by considering figure 4) Case1: Node Failure: if a node fails during communication in network. Let us suppose node fails. tep 1: As node will not able to get ack within its RTT (Round Trip Time) then node will send packet one more time, even if it is not able to get an ack then it will assume that node has been failed. tep : Node will select its next least cost and will send packet to route. Further we consider another situation, if node fails, node will not able to get ack within RTT, then node will roll back because there is no other path exist other than node to reach to destination. Then node will follow path. Advantage of having buffer in network is that in case of node failure there is no need to re-establish the path and can simply access packet from buffers present at intermediate nodes. Case : Link Failure: when a link fails inside a network during communication between source and destination. Let us take an example when link fails. Link failure case is similar to node failure tep 1: let be source node and be the destination node. Node will follow its routing table and see that intermediate buffered nodes are in two different paths so it will follow a path with minimum cost path i.e. and send packet through this path. tep : As the packet reaches to node, there is a link failure so; node will not be able to get ack within RTT. o node will follow another path by seeing its routing table i.e. which is next highest least cost distance between,. IV. PERFORMANCE ANALYI We have evaluate performance analysis of BBR (Table 5) in terms of network performance parameters i.e. throughput, resiliency against node/link failure, network congestion and delay. These parameters are defined as: Throughput: throughput is defined as how fast we can actually send data through a network. Resiliency against node/link failures: how resiliency implemented in case of failures. Network congestion: how much load on network exists during transmission of data between source and destination? Delay: how long it takes for an entire message to completely arrive at destination. ince [1] follows two-node disjoint path, packets broadcast to the network and follows both paths concurrently to reach the destination. In case when any intermediate node/link fails, then packets will automatically switch to unaffected path due to which load on that path increases and throughput decreases. In ROMER [1] throughput is higher

6 than resilient multicast and all packets follow single path. In the proposed approach buffers are placed according to least cost path selection and during packet transferring,it choose path which has less number of intermediate buffers from {, D}, so throughput increases because it sends data according to least cost path. In resilient multicasting packets [1] flows parallel and concurrently through -node disjoint path, in case of any node/link failure packets will shift to single path due to which cost of resilient multicast increases. But in ROMER [1] packets follows single path and forward packets according to credit limit. In case when initial credit limit is less, cost at each node is very high and number of failures exists in network, there are more chances of discarding packets by the node. In the proposed approach we employ buffers for storing the packets coming from previous nodes. In case any nonbuffered node fails, then we can get data from its preceding buffer and if buffered node fails then we get data from previous buffered node due to which re-establishment cost decreases. TABLE V. PERFORMANCE ANALYI OF BBR Parameters Performance Reasons Throughput Increases Because of least cost buffered path selection during transmission process. Resiliency against node/link failure Network congestion Increases Decreases Because transmission starts from previously buffered node. Due to limited buffer capacity, number of packets in network are less. Delay Decreases Due to transmission of packets (after transmission of first packet) Reliability and robustness Increases Because of implementation of resiliency, network is more reliable, chances of network failure are less C D E F C D E F C D E F C D E F C D E F C D E F C D E F C D E F Fig. 4. Buffer Allocation using BBR. V. CONCLUION In this paper we have proposed a resilient Buffer Based Routing Approach which ensures effective communication in WMN. We have allocated buffers according to minimum cost path into network that provides shortest path to reach destination. Each node maintains its own routing table which keeps additional information about buffered nodes in the network. We provide shortest path into network without using any shortest path routing algorithms. In the proposed BBR approach buffers are allocated alternatively which reduces reestablishment cost in network against node/link failures. We have shown comparisons between previous proposed approaches (resilient Multicast and ROMER) and the proposed BBR approach. BBR provides high throughput, reduces network congestion during node/link failures and ensure robustness into network. Further practical implementation of BBR approach for wireless mesh network is to be analyzed for accurate and absolute results. REFERENCE [1] Akyildiz, Ian F., and Xudong Wang. "A survey on wireless mesh networks."communications Magazine, IEEE 4.9 (005): -0. [] Trivedi, Kishor., Dong eong Kim, and Rahul hosh. "Resilience in computer systems and networks." Proceedings of the 009 International Conference on Computer-Aided Design. ACM, 009.

7 [] Jaggi, idharth, et al. "Resilient network coding in the presence of byzantine adversaries." INFOCOM th IEEE International Conference on Computer Communications. IEEE. IEEE, 007. [4] chaeffer-filho, Alberto, et al. "A framework for the design and evaluation of network resilience management." Network Operations and Management ymposium (NOM), 01 IEEE. IEEE, 01. [5] Yu, Yue, et al. "An adaptive approach to network resilience: Evolving challenge detection and mitigation." Design of Reliable Communication Networks (DRCN), 011 8th International Workshop on the. IEEE, 011. [6] Fick, David, et al. "A highly resilient routing algorithm for fault-tolerant NoCs."Proceedings of the Conference on Design, Automation and Test in Europe. European Design and Automation Association, 009. [7] Lee, anghwan, et al. "Proactive vs reactive approaches to failure resilient routing." INFOCOM 004. Twenty-third AnnualJoint Conference of the IEEE Computer and Communications ocieties. Vol. 1. IEEE, 004. [8] Rao,. iva Nageswara, Y K undara Krishna, and K. Nageswara Rao. "A urvey: Routing Protocols for Wireless Mesh Networks." International Journal of Research and Reviews in Wireless ensor Networks (IJRRWN) 1. (011). [9] Rakesh, Nitin, and Vipin Tyagi. "Failure recovery in XOR'ed networks." ignal Processing, Computing and Control (IPCC), 01 IEEE International Conference on. IEEE, 01. [10] upta, Bhagvan Krishna, Ankit Mundra, and Nitin Rakesh. "Failure Detection and Recovery in Hierarchical Network Using FTN Approach." (01). [11] Rakesh, Nitin, and Vipin Tyagi. "Failure Detection using Contour Approach on Network Coded Parallel Networks." Procedia Engineering 8 (01): [1] Zhao, Xin, et al. "Resilient multicasting in wireless mesh networks." Proc. 1th International Conference on Telecommunications [1] Yuan, Yuan, et al. "ROMER: resilient opportunistic mesh routing for wireless mesh networks." IEEE Workshop on Wireless Mesh Networks (WiMesh). Vol