PRIME: A partial path establishment based handover management technique for QoS support in WiMAX based wireless mesh networks

Computer Networks xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Computer Networks journal homepage: www.elsevier.com/locate/comnet

PRIME: A partial path establishment based handover management technique for QoS support in WiMAX based wireless mesh networks L. Rajya Lakshmi ⇑, Vinay J. Ribeiro, B.N. Jain Department of Computer Science and Engineering, IIT Delhi, New Delhi 110016, India

a r t i c l e

i n f o

Article history: Received 19 May 2014 Received in revised form 11 February 2015 Accepted 17 March 2015 Available online xxxx Keywords: WiMAX Wireless mesh networks Handoff QoS Crossover node Markov model

a b s t r a c t In this paper, we propose a novel handover management technique called PaRtIal path establishment based handover Management tEchnique (PRIME) for WiMAX based wireless mesh networks (WMNs). Different from the currently existing methods for WiMAX networks, PRIME addresses handover management in WiMAX WMNs deployed with distributed scheduling. In these networks, to continue the quality of service (QoS) constrained flows to a mobile node (MN) after its handover, a new path with the required bandwidth and which passes through its new base station (BS) needs to be established as quickly as possible. To address that issue, PRIME handles re-routing and scheduling issues of a handing over MN together. To provide lossless and seamless service, PRIME tries to establish new path(s) in the wireless mesh with the required bandwidth to the MN before it enters into the coverage area of the new BS. The present paper proposes a novel crossover node based partial path establishment algorithm to establish new path(s) which support QoS requirements of handoff calls. To analyze the performance of PRIME, the present paper proposes a multi-dimensional Markov model. Unlike previous models which analyze the performance of wireless networks, our proposed model represents nodes in terms of the number of transmission and reception available slots. The theoretical upper and lower bounds on the call dropping probabilities of handoff calls are obtained. To study the performance advantages of PRIME, we devise another handover management method called RFPHMT which does not use the concept of crossover base station in the new path establishment of a handing over node. The performance of PRIME and RFPHMT are compared in terms of call dropping probabilities and call setup delays. PRIME shows superior performance than RFPHMT. For a random topology, at a high call arrival rate of 1/2000 (calls/milliseconds), the handover call dropping probability of PRIME is 40% less than that of RFPHMT. The call dropping probabilities of PRIME with the simulations are always within the theoretical bounds which proves that the obtained bounds are close to the real call dropping probabilities. Ó 2015 Elsevier B.V. All rights reserved.

1. Introduction Wireless mesh networks (WMNs) are a promising technology to deliver various emerging wireless services to end ⇑ Corresponding author. E-mail address: [email protected] (L. Rajya Lakshmi).

users in a cost-effective manner. They are scalable, reliable and address the high bandwidth needs of current and future wireless applications. Two types of nodes exist in these networks: mesh routers/base station (MRs/BSs) and mesh clients [1]. MRs/BSs establish a wireless mesh among themselves which acts as a backbone of a WMN. The coverage area of a WMN can be extended by adding more MRs/

http://dx.doi.org/10.1016/j.comnet.2015.03.013 1389-1286/Ó 2015 Elsevier B.V. All rights reserved.

Please cite this article in press as: L. Rajya Lakshmi et al., PRIME: A partial path establishment based handover management technique for QoS support in WiMAX based wireless mesh networks, Comput. Netw. (2015), http://dx.doi.org/10.1016/j.comnet.2015.03.013

2

L. Rajya Lakshmi et al. / Computer Networks xxx (2015) xxx–xxx

Fig. 1. A sample network diagram to explain notations.

BSs to the existing network. Mesh clients access services by getting connected to one of the MRs/BSs. They can be easily deployed and used in emergency situations, disaster affected areas etc. to mention a few. A wireless backbone is an important part of a WMN. One of its responsibilities is the backhaul networking to deliver quality of service (QoS) constrained services to users while addressing fairness issues. Till now many WMNs are deployed worldwide and most of them are IEEE 802.11 based WMNs [27,28]. Analysis in previous work [26] revealed that the IEEE 802.11 MAC is not suitable for the backhaul networking. In contrast, IEEE 802.16 technology, which consists of TDMA (time division multiple access) MAC, distributed scheduling among the nodes, more capacity and capability to provide fairness to network flows, is a promising technology for backhaul networking in WMNs. WiMAX WMN architecture considered in this paper (see Fig. 1) consists of a wireless mesh among the WIMAX base stations; clients can access the network services by connecting to one of the BSs. Two types of scheduling methods can be used in WiMAX mesh networks: centralized and distributed. With centralized scheduling, the BS is responsible for all scheduling tasks. With distributed scheduling, nodes coordinate with their two-hop neighbors to come up with a collision free schedule. In this paper, we consider WiMAX WMNs deployed with distributed scheduling. Handover management is an important research issue related to any wireless network. Even though many mobility management methods are developed for other wireless networks, the handover management in TDMA WiMAX WMNs deployed with distributed scheduling has not been addressed before. The present paper proposes a novel method for handover management in these networks. In TDMA WiMAX WMNs, data transmission between a source S-MN (see Table 3 for the list of notations) and a destination D-MN (a flow) takes place through a multihop path between them. During the establishment of the flow, the required bandwidth is reserved at all the nodes along a

multihop path between S-MN and D-MN. Assume that the destination node D-MN (see Fig. 1) handovers from the coverage area of its currently serving BS (D-BS) to the coverage area of some other BS (T-BS). In order to continue the QoS-constrained1 flow between nodes S-MN and D-MN after D-MN’s handover, a new route (Rn ) from S-MN to D-MN needs to be established before the handover of D-MN. Rn should pass through T-BS and the required bandwidth must be available along that path. To achieve that, one simple solution is to start new path establishment from the source S-MN. In that case, the route re-discovery delay might cause interruption to the services flowing to D-MN. In addition to that, some packets may get lost during the handover process. Hence, we need an algorithm which finds new path(s) for D-MN with the required bandwidth as quickly as possible. To address that issue, this paper proposes a crossover node based partial path establishment algorithm for TDMA WiMAX WMNs. There are many research papers that discuss various issues related to WiMAX WMNs deployed with distributed scheduling. Various scheduling and call admission control methods are developed for TDMA WiMAX WMNs. A survey on these methods is given in [29]. But none of these works addresses end-to-end QoS guarantees to flows. An analysis [30] of the research activities by the standardization groups reveals that the IEEE 802.16 based TDMA WMNs are still active in the research community. Hence it is very important to address the issue of preservation of end-to-end QoS requirements of the flows of handover nodes. The partial path establishment method proposed in this paper tries to utilize the old path of an MN after its handover. This is the first novel quality of the proposed method. It tries to find a node which is on the old path and which has a route to the target BS of the MN with the required bandwidth. Such a node is called a crossover node of the old path and new path of the MN. In Fig. 1, 1 QoS-constrained flow requires certain bandwidth allocation along its path.

Please cite this article in press as: L. Rajya Lakshmi et al., PRIME: A partial path establishment based handover management technique for QoS support in WiMAX based wireless mesh networks, Comput. Netw. (2015), http://dx.doi.org/10.1016/j.comnet.2015.03.013

L. Rajya Lakshmi et al. / Computer Networks xxx (2015) xxx–xxx

assume that a flow has been established between S-MN and D-MN along the path ‘‘Old Path’’. To handle handover of D-MN, a new path to D-MN through T-BS is found that is ‘‘New Path-1’’. Instead of finding a brand new path from S-MN to D-MN through T-BS, by finding a suitable crossover node (node C in this case) and a QoS-providing partial path from that crossover node to T-BS, a new path can be established for S-MN to D-MN flow quickly. Two advantages of finding a crossover node are: (i) A new path can be established to the MN quickly. (ii) Packet delays and packet drop rate can be reduced. These performance advantages of the proposed method are established through simulation experiments and discussed in Section 7. The proposed method does not start the new path establishment process from the source of the flow of a handover MN. It searches for the crossover nodes on the old path. These crossover nodes must have paths to T-BS. There might be many such crossover nodes. The proposed method, by starting the new path establishment process from T-BS, searches a crossover node that is nearest to TBS. As a result it would be able to establish new paths for handover mobile nodes quickly and support the QoS requirements of their flows. In the situations where no new path with the total required bandwidth is available for a handing over MN, the proposed method establishes multiple paths which can together provide the required QoS to the MN. This is the second novel quality of the proposed method. Assume that, in the network shown in Fig. 1, ‘‘PP-1’’ can provide only 4 of 6 slots required for the flow from S-MN to D-MN and another path ‘‘PP-2’’ can provide the remaining 2 slots. By considering node D as a crossover point, paths ‘‘PP-1’’ and ‘‘PP-2’’, together fulfill the QoS requirements of the established flow. In this paper, we propose a multidimensional Markov model to analyze the performance of the proposed partial path establishment algorithm. In the literature Markov chains are used to analyze the performance of various types of networks including WiMAX networks [4,23,24]. In most of those works, the state of the BS is represented in terms of the number of currently admitted calls of various service classes [4]. Also, they do not distinguish between the transmission available and reception available slots. Based on the number of admitted calls, the number of free slots are estimated. In contrast, the model developed in this paper represents the state of a node/BS in terms of the number of transmission available and reception available slots. With this representation, one can know the number of transmission available and the number of reception available slots at a node/BS directly from its state; one need not estimate free slots based on the number of admitted calls. The main contributions of this paper are: (1) A novel handover management method, PRIME, for WiMAX mesh networks. Different from the previous handover management methods proposed for WiMAX networks, PRIME addresses the handover

3

management in TDMA based WiMAX WMNs deployed with distributed scheduling and handles routing and scheduling issues of a moving MN together. (2) Statistical analysis of the proposed method using a multi dimensional Markov model to obtain upper and lower bounds on the handover call dropping probabilities. The rest of the paper is organized as follows. Section 2 provides a brief explanation of the mobility management methods and Markov models developed in the literature. Section 3 discusses the architecture of the WiMAX WMNs, routing and scheduling algorithms used. Section 4 explains our approach towards handover management in WiMAX WMNs. Section 5 explains a handover management method that uses Race Free routing Protocol (RFP) to establish new paths to handover MNs. The performance of PRIME is compared with the performance of this method. Section 6 discusses the multi-dimensional Markov model which analyzes the performance of the proposed handover management method. Section 7 discusses simulation results. A performance analysis of PRIME and RFPHMT as well as a comparative study of the simulation and analytical results of PRIME are given in this section. Section 8 concludes the paper. 2. Related work This section discusses some of the related research works starting with methods for rerouting in wired backbones of point-to-multipoint (PMP) networks. Later, it describes handoff management methods proposed in the literature for WiMAX and WiFi WMNs. Finally, it discusses some performance analysis methods which use Markov chains. 2.1. Handoff in multihop networks There are two fundamental types of handoffs in wireless networks: hard handoff and soft handoff. In a hard handoff, a mobile node (MN) first breaks its connection with the old base station and then establishes a new connection with the new base station whereas in the soft handoff, an MN communicates simultaneously with a set of base stations called an ‘‘active set’’ and typically selects the BS with the best signal strength from the active set as the target BS. A semi soft handoff can be described as a subcategory of soft handoff, in which a new route to the MN through a new BS is found before the handoff of the MN, to reduce its handoff latency and packet loss. 2.1.1. Rerouting in wired backbone of point-to-multipoint networks In the literature, some previous works addressed the partial handover rerouting problem in ATM networks [25]. In [19], the authors proposed an incremental path re-establishment based hybrid handover protocol for local area ATM networks. The network architecture they have considered has a wired mesh among the ATM switches.

Please cite this article in press as: L. Rajya Lakshmi et al., PRIME: A partial path establishment based handover management technique for QoS support in WiMAX based wireless mesh networks, Comput. Netw. (2015), http://dx.doi.org/10.1016/j.comnet.2015.03.013

4

L. Rajya Lakshmi et al. / Computer Networks xxx (2015) xxx–xxx

The proposed scheme finds a nearest common ATM switch which falls on the old path and has a route to the new cluster switch to handle the handover of an MN. The crossover switch/node is always the same for a given pair of switches. Different from this work, the present paper handles handover management in TDMA WiMAX mesh networks where the mesh among the BSs is wireless. In addition, the crossover node for a given pair of BSs might be different based on the changing traffic and environmental conditions. In [18], a mobility management scheme was proposed to support fast handoff in a wireless ATM local area network (LAN). After the handover of an MN, the cells of all connections of the MN are forwarded from the new serving portable base station (PBS) to the old serving PBS using provisioned virtual paths. It is not mentioned how the required bandwidth is allocated on the link between the new serving PBS and the old serving PBS and the course of action in case bandwidth is not available. Different from this method, in PRIME, the new paths with the required bandwidth are established to the handing over nodes dynamically as the requirement arises. Since, in WiMAX WMNs, as the flows enter and leave the network, the bandwidth availabilities along various paths change. If, we reserve bandwidth on paths between BSs in advance specifically to transmit handover traffic only, then we will be wasting bandwidth, as it may be used infrequently. Fast Intra-Network and Cross-Layer Handover (FINCH) [15] is proposed to reduce the packet drop rate and packet delivery delay in mobile WiMAX PMP networks. In this method also, control messages are used to establish new paths after handovers. In [11], the gateway router sends the traffic of a mobile node to both the old BS and new BS to reduce the packet loss during mobile WiMAX handover. In their network architecture the BSs are connected through wired links.

2.1.2. Handoff in WiMAX WMNs In [20], authors have modified the mobility management method given in the IEEE 802.16 standard to develop a seamless mobility management method for the WiMAX WMNs but did not address the issue of establishing a new path from source to destination. They did not mention how the required bandwidth is allocated along the new path of a handover MN. In their method, the undelivered packets of a handover MN are buffered at the serving BS instead of at the gateway. After MN’s handover, the serving BS sends buffered packets to the target BS. In [21], a backtracking based rerouting algorithm is proposed for WiMAX WMNs which needs the knowledge of optimal routes from every subscriber station to the others. In this method, each subscriber station transmits location update messages to obtain local information of the network. These location update messages are used to construct a new route for an MN which is handing over from the coverage area of one subscriber station to the other. This method leads to packet flooding in the network for certain duration. Different from this method, PRIME does not require knowledge of such optimal routes. It finds new paths to handover nodes as the requirements arises.

In [22] layer-3 handover is performed prior to the layer2 handover (based on the handover anticipation information). This policy is adopted to reduce the packet drops and handover latency involved in the mobility management method given in IEEE 802.16e [9] where the layer-3 handover is performed after the completion of layer-2 handover. It is not enough to find a new route for a handing over MN; that path should have enough bandwidth to continue the QoS-constrained flows that are passing through the MN. The PRIME method proposed in this paper addresses that issue. Most of the distributed scheduling based WiMAX mesh deployments, handle routing and scheduling separately. Whenever an MN performs handover, a new path is identified first. Then the packets are transmitted along the identified path by hop-by-hop routing. Nodes on the new path, after receiving the packets of the flow, use distributed scheduling messages to allocate bandwidth to their next hops on the new path. It might happen that the required bandwidth is not available between two nodes along the new path. In that case it is difficult to support the QoS requirements of the flows of handover nodes. To address that issue, PRIME handles routing and scheduling issues together to support the QoS requirements of the flows of handover nodes at the required levels. 2.1.3. Handoff in WiFi WMNs In the literature many mobility management methods are proposed for the IEEE 802.11 based WMNs in which MAC layer protocol uses CSMA/CA mechanism. These methods can be categorized as: routing based, tunnel based and multicast based [5]. I-mesh [12] and MEMO [13] are two methods which come under the category of routing based methods where a routing algorithm is used to establish new routes to the MNs after their handover. In the tunnel based methods, a tunnel is established between the old-BS and the new-BS to reduce the packet drops. Mesh Mobility Management [7] and Ant [14] are the methods that come under this category. In S-mesh [2], a multicast based method, the BSs which are in the transmission range of an MN form a group called client control group and the BSs with the best connectivity to that MN form a client data group (CDG). During handover of the MN, the packet loss rate is almost nil since the target BS is always in the CDG. In contrast to these methods, PRIME is proposed for handover management in TDMA based WiMAX WMNs. The handover management methods proposed for WiFi WMNs cannot be used for WiMAX WMNs, since the carrier access methods of WiMAX and WiFi networks are different. In addition, none of the method discussed above addresses routing and scheduling issues of a handover MN together. To support the QoS requirements of handoff calls, PRIME handles routing and scheduling issues of handover nodes together. 2.2. Performance analysis of wireless networks using Markov chains In [4], a Markov chain analysis is given to analyze the impact of multiple channels and multiple radios on the performance of TDMA based mesh networks. Even though

Please cite this article in press as: L. Rajya Lakshmi et al., PRIME: A partial path establishment based handover management technique for QoS support in WiMAX based wireless mesh networks, Comput. Netw. (2015), http://dx.doi.org/10.1016/j.comnet.2015.03.013

L. Rajya Lakshmi et al. / Computer Networks xxx (2015) xxx–xxx

flows with different priorities are considered, their bandwidth requirements are assumed to be the same, that is, one slot per frame. The state of the system is represented in terms of the number of calls of each service class on each channel. Based on the number of admitted calls, the number of free slots are estimated approximately. Their model does not distinguish between the transmission available and reception available slots. Different from this model, in the Markov model proposed in this paper, the state of a node is represented in terms of the number of transmission and reception-available slots at the node. Markov chains are also used for the performance analysis of other networks. An improvement in the reduced call blocking probability through integrated Cellular and ad hoc Relaying (iCAR) is studied in [24] using Markov chains. In this modeling, the state of a cell is represented in terms of the number of busy channels. The call acceptance ratio of TWiLL (Throughput enhanced Wireless in Local Loop) is analyzed in [23]. All these methods analyze various performance parameters of cellular networks, but the Markov model proposed in this paper analyzes the dropping probabilities of handover calls in TDMA based WiMAX WMNs. Also, as described in Section 1, the way we represent the state of a BS is different from the way the above mentioned methods represent the state of a cell. 3. System architecture A WiMAX based wireless mesh network can be formed of a set of WiMAX base stations and different sets of mobile nodes under the coverage area of each base station. The BSs establish a wireless mesh network among themselves dynamically. That mesh acts as the backbone for the network. In WiMAX WMNs time is divided into fixed length frames, and each frame consists of a control subframe and a data subframe [3]. WiMAX mesh supports two types of scheduling: centralized scheduling and distributed scheduling. BSs perform all scheduling tasks in the centralized scheduling. They receive bandwidth requests from all subscriber stations, allocate bandwidth and distribute bandwidth allocation information in the network. WiMAX supports two types of distributed scheduling: coordinated and uncoordinated. In the coordinated distributed scheduling, there is no possibility of packet collision, where as in the case of uncoordinated distributed scheduling, the control packets transmitted by two nodes may collide. For data transmission, each node in the network coordinates with its two-hop neighbors to come up with a collision free schedule. In this paper we are using coordinated distributed scheduling to establish end-to-end paths with the required bandwidth. 3.1. Coordinated distributed scheduling With the coordinated distributed scheduling, data transmission is facilitated by a three-way handshake procedure and with the help of three information elements (IEs) of the mesh distributed schedule (MSH-DSCH) message. These information elements are: request IE, grant

5

IE, confirmation IE. When a node S wants to send data to one of its one-hop neighbors, say node D, it selects a set of slots to transmit to D and sends a request IE r 1 in its MSH-DSCH opportunity to node D. Node D, after the reception of r1 , sends a grant IE to node S with a set of slots allocated for data transmission from node S to node D. After the reception of grant IE from node D, node S sends a confirmation IE to node D. This procedure by itself does not establish end-to-end QoS providing paths in WiMAX WMNs. The fourth information element that can be transmitted through MSH-DSCH messages and helps to avoid interference is the availability IE. It can be used by the nodes to inform their two-hop neighbors about slot allocations which have taken place for data transmission and reception. 3.2. Routing and scheduling algorithm In WiMAX WMNs, when a node S wants to send some real time data to some other node D, a QoS (bandwidth) providing multi-hop path needs to be established between S and D. We require a routing and scheduling algorithm for that purpose. Race Free routing Protocol [10] is a routing algorithm to establish end-to-end QoS providing routes in TDMA based ad hoc networks. We have extended this algorithm to WiMAX WMNs and used it as the underlying routing and scheduling algorithm to establish QoS providing multi-hop paths. A brief description of RFP algorithm which is very essential to understand the proposed PRIME is given below. Race Free Protocol RFP algorithm uses QREQ and QREP messages in the QoS (bandwidth) providing path establishment process. According to this protocol, whenever a QoSconstrained flow is to be established between two nodes, the source finds out whether the required bandwidth is available to at least one of its one-hop neighbors. If enough bandwidth is available, then broadcasts a QREQ to all its neighbors. The nodes which receive the QREQ follow the same procedure as that of the source to forward the QREQ. As the QREQ traverses from the source node to the destination, it allocates the required bandwidth temporarily for certain timeout period. If the QREQ reaches the destination, then the destination and all nodes along the path from the source to the destination reserve the previously allocated slots for source–destination flow for the life time of the flow. In RFP, every node maintains a routing table to record which flows are passing through it. Nodes maintain information regarding allocated slots and reserved slots in separate tables. Also, when a node either allocates or reserves certain slots for some flow, it broadcasts this information in its two-hop neighborhood to avoid interference. This paper integrates the coordinated distributed scheduling of WiMAX mesh mode and RFP to develop a routing and scheduling algorithm for the proposed architecture of WiMAX WMNs. QoSREQ and QoSREP are exchanged between the nodes through MSH-DSCH messages in the MSH-DSCH transmission opportunities. Nodes use QoSREQ and QoSREP for the initial path establishment. The structures of QoSREQ and QoSREP are the same as those of QREQ and QREP of RFP algorithm

Please cite this article in press as: L. Rajya Lakshmi et al., PRIME: A partial path establishment based handover management technique for QoS support in WiMAX based wireless mesh networks, Comput. Netw. (2015), http://dx.doi.org/10.1016/j.comnet.2015.03.013

6

L. Rajya Lakshmi et al. / Computer Networks xxx (2015) xxx–xxx

respectively. Also nodes follow the same procedure as that of RFP to establish QoS providing paths using these messages. 3.3. Data structures In the proposed architecture, every node is given a two digit identification number. Various messages used in the proposed algorithm are identified by the pair of (source ID, sequence number). Important data structures maintained by each base station are: routing table, neighborhood table, slot allocation tables and Mobile Node Information for Handover table (MNIH table). Routing table: Similar to RFP method, each entry of the routing table stores source ID, destination ID, next hop on the path from the source to the destination, and slots in which data is to be received from the previous node on the path, and slots in which received data is to be transmitted from this node to the next hop. Neighborhood table: In this table, nodes store paths to their two-hop neighbors along with their bandwidth availabilities. A more detailed explanation of the use of this table is given in the next section. Slot allocation tables: At a node, each slot of the data frame could be in one of three states: free, allocated, reserved. Separate tables are maintained at each base station to store information regarding the slots allocated or reserved for transmissions and slots allocated or reserved for receptions. All allocated slots at a node X are associated with a timeout period. By the end of this timeout period if these slots are not reserved, they will be freed by node X. MNIH table: In this table, the BS maintains information about all client nodes currently associated with it. Important fields of each entry of this table are: node ID of a client, if that client is mobile, the target BS (T-BS) into whose coverage area that client is moving, and the time of entry into the coverage area of T-BS, crossover node information and so on. 3.4. Slot allocation rules To avoid interference, all nodes in the network follow three rules while allocating or reserving slots for their neighbors. Node A can allocate a slot S to one of its onehop neighbors, say node B, for data transmission between A and B, if and only if the following three rules are true: (1) node A and node B are neither receiving nor transmitting in slot S, (2) No one-hop neighbor of node A is receiving or transmitting in slot S, (3) No one-hop neighbor of node B is transmitting in slot S [4]. 4. Partial path establishment based handover management technique for WiMAX WMNs The proposed method consists of two parts: partial path establishment algorithm (Section 4.1) and handoff algorithm (Section 4.2). In Fig. 2, in order to continue QoS-constrained service to D-MN after its handover, the route re-discovery algorithm tries to find a QoS-providing path to D-MN through T-BS. If this new path establishment

Fig. 2. An example network diagram.

starts from S-BS, the route re-discovery delay might be high and causes interruption to the service flowing to D-MN. In addition, some packets may get lost during the handover process. To avoid these two problems we developed a partial path establishment algorithm (PPEA) as a part of the proposed handover management method. Some important observations of partial path establishment are mentioned below with the help of a concept of crossover node. A crossover node is a node (BS) which is on the old path and has a QoS-providing route to T-BS. (1) By finding a crossover point which is on the old route of a mobile node and has a route with the required bandwidth to the new BS of the mobile node, we expect to establish a partial path with required bandwidth quickly, if there exists one such path. When D-MN1 shown in Fig. 2 handovers to node C, by finding a suitable crossover node (node A in this case) and a QoS-providing partial path from that crossover node to node C, one can finish the new path establishment to D-MN1 through node C quickly. (2) If it is not possible to find a single partial path from a crossover point to the target base station with the required bandwidth, then by finding multiple partial paths to the target base station, which can together fulfill the bandwidth needs, the required QoS can still be provided to D-MN. Suppose in the network shown in Fig. 2, ‘‘PP-1’’ can provide only 4 of 6 slots required for the flow from S-MN1 to D-MN1 and another path ‘‘PP-2’’ can provide the remaining 2 slots. By considering node D as a crossover point, paths ‘‘PP-1’’ and ‘‘PP-2’’, together can fulfill the QoS requirements of the flow between S-MN1 and D-MN1.

4.1. Partial path establishment algorithm PPEA, consists of 5 phases. Before we move onto the detailed descriptions of those phases some assumptions are given below. Assumptions: (1) T-BS, into whose coverage area D-MN is entering and the time of the entrance of D-MN into the coverage area of T-BS is known to D-BS. Methods are available in the literature to find the destination location of a

Please cite this article in press as: L. Rajya Lakshmi et al., PRIME: A partial path establishment based handover management technique for QoS support in WiMAX based wireless mesh networks, Comput. Netw. (2015), http://dx.doi.org/10.1016/j.comnet.2015.03.013

L. Rajya Lakshmi et al. / Computer Networks xxx (2015) xxx–xxx

mobile node and time to reach that location (based on the speed and direction of the mobile node). (2) All mesh routers (BSs) and mesh clients use the same channel for control and data traffic transmission. PPEA, based on these assumptions, before the handover of a mobile node, pre-computes either a single route from some crossover point to T-BS with the required bandwidth or a set of routes from a crossover point to T-BS which together can provide the required bandwidth. 4.1.1. Phase 1: Initialization of new path establishment process Sometime (which varies with the number of control slots available in the control subframe) before the entrance of D-MN into the coverage area of T-BS, D-BS asks T-BS to establish new route/routes to handle future handover of D-MN by sending a PPEreq IE. While sending the PPEreq IE to T-BS, D-BS sends the old path (from S-MN to D-MN) and the required bandwidth information also. The new route establishment takes place in 4 stages. 4.1.2. Phase 2: Search for a crossover node in the neighborhood Upon the reception of PPEreq, T-BS tries to find a crossover node for D-MN in its two-hop neighborhood. Every node maintains paths to its two-hop neighbors along with the bandwidth availabilities in a table called neighborhood (NH) table. T-BS checks for the existence of a crossover node in its two-hop neighborhood. If this phase succeeds then there is no need to execute the other phases of the PPEA. The success rate of this phase depends upon the number of hops that are considered for the neighborhood computation. In the case of failure of this phase the remaining phases will be executed. For example, in the network shown in Fig. 2, assume that a flow is established between S-MN1 and D-MN1 along the path S-MN1, G, D, A, B, E, D-MN1 and the required bandwidth of 6 slots is reserved along that path. Table 1 shows the NH table of node C. After receiving the PPEreq from node E, by checking its NH table, node C identifies node A as the required crossover point; the required bandwidth is available from node A to node C. Node C reserves the required bandwidth along the partial path from A to C for the flow between S-MN1 and D-MN1 and sends this crossover node information to D-BS, that is node E, through a PPErep IE.

Table 1 Neighborhood of T-BS (Node C). Node

Path

Bandwidth

A M F E B B H I D D L

C-A C-M C-F C-E C-E-B C-A-B C-E-H C-F-I C-A-D C-M-D C-M-L

6 5 4 2 2 4 2 3 2 3 2

slots slots slots slots slots slots slots slots slots slots slots

7

4.1.3. Phase 3: Crossover node search by transmitting a request In case T-BS fails to find a crossover node in its two-hop neighborhood, T-BS tries to establish a new path to D-MN by broadcasting a pre-handover request PHOQREQ IE. The structure of PHOQREQ IE is similar to that of QoSREQ IE, but it carries some extra fields to facilitate the crossover node identification. T-BS tries to allocate (for reception) the bandwidth required for S-MN to D-MN flow to its one-hop neighbors. However, to some of its neighbors, it would be able to allocate only a fraction of the required bandwidth. T-BS allocates different sets of slots to its neighbors to establish multiple paths in case no single path with the required bandwidth is available for D-MN. Then T-BS enters information about slot allocation onto the PHOQREQ and broadcasts the same in its neighborhood. Node X, a neighbor of T-BS, to which T-BS allocated some bandwidth, upon reception of the PHOQREQ, allocates its respective slots mentioned in the PHOQREQ for the data transmission. If it is not a crossover node, then follows the same procedure as that followed by T-BS to forward the received PHOQREQ; otherwise executes phase 4 to send a reply to the received request. But the neighbors of T-BS to whom only a partial amount of the required bandwidth is allocated, consider that partial bandwidth as the required bandwidth while executing the forwarding process for the received PHOQREQ. Eventually, a copy of PHOQREQ reaches some crossover node. In the case of network shown in Fig. 2, assume that a flow with the bandwidth requirement of 6 slots has been established between S-MN and D-MN along the path S-MN, D, A, B, E, H, D-MN. Assume that, to establish a new path to D-MN, T-BS allocates 6 slots along I-F link but allocates only 4 slots along the links I-J and I-H. In that case, nodes J and H try to allocate 4 slots to their one-hop neighbors before broadcasting PHOQREQ in their neighborhoods. 4.1.4. Phase 4: Bandwidth reservation along the partial path Assume that a crossover node receives the first copy of the PHOQREQ (transmitted by T-BS) and that message has allocated the total required bandwidth along the path to TBS. Then that crossover node reserves the bandwidth for SMN to D-MN flow and sends a reply (PHOQREP) to respond to the received PHOQREQ and to reserve the previously allocated slots along the path to T-BS. If a crossover node receives the first copy of the PHOQREQ and that message has allocated less bandwidth than the required for the flow from S-MN to D-MN, then it checks for the other links along which it may receive the other copies of the same request. If there are no other links remain, then the crossover node drops the first copy as it is not possible to establish path/paths with the required bandwidth. Otherwise the crossover node waits for certain amount of time. By the end of the waiting time or after receiving requests along all other links, if it receives a request with all the requested bandwidth allocated, then it sends the reply (PHOQREP) to that request only and drops the other copies; otherwise it sends replies (PHOQREPs) to a set of requests which together fulfill the bandwidth requirement of the flow from S-MN to D-MN and drops the other copies.

Please cite this article in press as: L. Rajya Lakshmi et al., PRIME: A partial path establishment based handover management technique for QoS support in WiMAX based wireless mesh networks, Comput. Netw. (2015), http://dx.doi.org/10.1016/j.comnet.2015.03.013

8

L. Rajya Lakshmi et al. / Computer Networks xxx (2015) xxx–xxx

After sending the PHOQREP, the crossover node starts sending the data of the respective flow to both D-BS and T-BS. The crossover node continues this bisection of data of D-MN until it receives a Stop Bisection (SPB) message from T-BS. T-BS starts buffering the received data of D-MN to handle its future handover. The multi-path establishment strategy definitely improves the QoS provisioning capability of PRIME, but it also involves some overheads. The crossover node has to break packets into fragments. The sizes of these fragments are proportional to the data rates provided by the paths involved in the multi-path data transmission. In case multiple paths are established from a crossover node to the target base station of an MN, packets of the respective flows of MN are split at the network layer. Crossover base station fragments the packets of a flow of the MN in proportion to the data rates provided by the multiple paths. Target base station is responsible for collecting various fragments of a packet and assembling them. It buffers the fragments that are received with lesser delays, waits for the other fragments of the corresponding packet, assembles the packet and sends the same to the corresponding mobile node. Since, the crossover and target BSs are static and do not have any power constraints, the increased overheads are not an issue. 4.1.5. Phase 5: Conclusion of new path establishment process After receiving the PHOQREP/PHOQREPs, T-BS reserves the previously allocated slots for the data reception. It also sends the crossover node information to D-BS using a PPErep IE. D-BS enters this crossover node information of D-MN into its MNIH table. If T-BS receives replies from some other crossover node, it sends a PHOQCANCEL IE to cancel the path established from that crossover node to itself. If, instead of D-MN, S-MN performs handover from S-BS to some target BS T-BS1, above explained procedure can be used by S-MN, S-BS, and T-BS1 to establish a new path from S-MN to D-MN through T-BS1, and the handover protocol given below can be used by S-MN to perform handover.

Also, when compared with the scheme [19], in which the crossover nodes of various node pairs are fixed, the proposed partial path establishment process provides superior performance. In the fixed crossover node based methods, if the required bandwidth is not available along the new path from the crossover node to target BS of an MN, then the flow(s) of that MN need to be dropped. The proposed partial path establishment method avoids some of such cases as the crossover nodes and new paths are not prefixed according to PRIME. PRIME explores all available paths and if a path with the required bandwidth is available, then allocates the bandwidth along that path and helps in supporting the QoS requirements of the flow(s) of MN. 4.3. Handover protocol When D-MN wants to handover to T-BS, it first establishes mesh connectivity with T-BS and sends a HReq message to D-BS. After the reception of the HReq message, DBS tries to find the crossover node information of D-MN in its MNIH table. If it is able to find the crossover node information in that table, that means a QoS providing path/paths has already been established for D-MN through T-BS. In that case D-BS sends the sequence number of the last data frame it has successfully sent to D-MN through a Last Packet Information (LPI) message to T-BS and HRep message to D-MN. In contrast if D-BS fails to find the crossover node information of D-MN, it sends HRep to D-MN and asks S-BS to establish a new path to D-MN through T-BS and also cancels the old path from S-MN to D-MN using a QCANCEL message. D-MN starts the actual handover after the reception of HRep. D-MN terminates its connection with D-BS. T-BS, upon reception of the LPI message, informs the crossover node to stop sending traffic destined for D-MN to D-BS by sending a Stop Bisection (SPB) message. After the reception of SPB message, the crossover node cancels the part of old path from itself to D-BS using a QCANCEL message. The message flow diagram of this protocol is shown in Fig. 3.

4.2. Benefits of PRIME PRIME starts a new path establishment of an MN sometime before the handover of MN. If a new path with the required bandwidth is available, then PRIME allocates the required bandwidth for the flow(s) of MN along that new path. By the time MN completes handover, new path is available with the required bandwidth. By doing this, PRIME would be able to reduce the packet delivery delays and packet drop rates of the flow(s) of MN. The other novel quality of PRIME is its partial path establishment process. With this process, the target BS of a handover MN starts new path establishment for the MN. This saves lots of control overhead when compared with the path establishment process in which the source starts new path establishment for a handover mobile node. For the source to start new path establishment, first the handover information in the form of a message has to travel along a long path to reach the source. Then the source starts the new path establishment for the flow(s) that are running to the MN.

Fig. 3. Message flow diagram of PRIME.

Please cite this article in press as: L. Rajya Lakshmi et al., PRIME: A partial path establishment based handover management technique for QoS support in WiMAX based wireless mesh networks, Comput. Netw. (2015), http://dx.doi.org/10.1016/j.comnet.2015.03.013

L. Rajya Lakshmi et al. / Computer Networks xxx (2015) xxx–xxx

9

transmission in a frame is M. The state of a node according to this Markov model is an approximation of the state of the node according slot allocation rules of PRIME. The state transitions explained in the following discussion are approximations of the actual state transitions. The number of slots available at a node changes if either of two events occurs: (i) a call request arrives at that node (ii) slot reservation takes place in its interference range. Now, we look at these two call arrival rates in detail. We assume that the call arrival rate at a node X is Poisson distributed with an exponential average kX O . But the distribution of calls originated at other nodes and passing through node X is not Poisson distributed due to the splitting of Poisson streams [4]. By using the Kleinrock’s Independence Assumption [17] we regard the cumulative call arrivals at node X is Poisson distributed with mean arrival rate of (kX ). The mean call arrival rate (kX ) at a node X is defined as:

kX ¼ Fig. 4. Message flow diagram of RFPHMT.

5. RFP based handover management method (RFPHMT) In the literature there are no handover management methods which handle handovers in TDMA WiMAX WMNs deployed with distributed scheduling. To study the performance advantages of PRIME which finds partial paths to handing over nodes, another handover management method is required which does not use crossover point in new path establishment of handing over nodes. So, we devise such a handover management method called RFPHMT in which the source finds fresh path for the handing over destination node. In this method, sometime (which varies with the number of control slots available in the control subframe) before the movement of D-MN into coverage area of T-BS, D-BS asks S-BS to establish a new path to D-MN through T-BS by sending a PPEreq message. After the reception of that message, S-BS using the RFP algorithm establishes a route to the T-BS with the required bandwidth. After the successful path establishment, S-BS sends a PPErep to D-BS and starts sending the data of D-MN to both D-BS and T-BS. The handover protocol of the RFPHMT is similar to that of PRIME. S-BS plays the role of crossover node in this handover protocol. The message flow diagram of RFPHMT method is shown in Fig. 4. 6. Probabilistic analysis This section discusses the proposed mathematical model which analyzes the performance of PRIME. A multi-dimensional Markov model is developed to analyze the performance of WiMAX mesh networks. The states of a node are represented in terms of the number of transmission available slots and the number of reception available slots at the node. A node is in state ði; jÞ means i slots are available for transmission and j slots are available for reception at that node. The initial state of a node is ðM; MÞ if total number of slots available for data

X ðf ði; XÞ  kio Þ þ kXO ;

ð1Þ

i

where f ði; XÞ is the fraction of calls originating from i and passing through node X and kio is the call arrival rate at node i. 1=lX is the call departure rate of node X. Now let us look at computation of f ði; XÞ. The destination of a call originated at node i is selected randomly. If the number of nodes in the network is n, then the probability of node X 1 being selected as the destination of the calls originated at node i is kio =n. There might be many paths, say n1 , between nodes i and X 1 . Among these paths, say n2 , paths pass through node X. The fraction of calls that originate at node i, end at node X 1 and pass through node X is: ðn2  kio Þ/ðn  n1 Þ. Similarly, such fractions for other destinations can be computed to obtain the cumulative fraction, that is, f ði; XÞ. The arrival rate of availability IEs at a node X from any arbitrary interfering node, node Y, is proportional to the call arrival rate at node Y. Node X receives similar availability IEs from all the nodes in its interference range. Interference call arrival rate (in terms of availability IEs) at a node X, denoted by kIX , is the summation of call arrival rates at all nodes in the interference range of node X. Based on the states of various slots at a particular time at a node there are ððM þ 1Þ  ðM þ 2ÞÞ=2 states possible for it; it could be in any one of the following states: ðM; MÞ; ðM  1; MÞ; . . . ; ð0; MÞ; ðM  1; M  1Þ; . . . ; ðM  2; M 2Þ; . . . ; ð0; 0Þ. If a node X is in the state ði; jÞ, then we denote node X as Xði; jÞ, its steady state probability is Q ði; jÞ, call arrival rate is ki;j , call departure rate is li;j and interference call arrival rate is kIi;j . We also assume that allocations for transmission and allocations for reception take place at node X with equal probability ðki;j =2Þ. Similarly, availability IEs for transmission and reception arrive at node X with equal probability ðkIi;j =2Þ. It is assumed that the following events, deallocation of a slot which is reserved for transmission, deallocation of a slot which is reserved for reception, arrival of availability IE with transmission deallocation and arrival of availability IE with reception deallocation, occur at node Xði; jÞ with equal rate li;j /4. In the following discussion, the state transitions of a node X which in state ði; jÞ are discussed.

Please cite this article in press as: L. Rajya Lakshmi et al., PRIME: A partial path establishment based handover management technique for QoS support in WiMAX based wireless mesh networks, Comput. Netw. (2015), http://dx.doi.org/10.1016/j.comnet.2015.03.013

10

L. Rajya Lakshmi et al. / Computer Networks xxx (2015) xxx–xxx

If i > 0 and j > 0, then an allocation of a slot at node Xði; jÞ for transmission changes its state to Xði  1; j  1Þ. Such allocations occur with the rate ðki;j =2Þ. If i ¼ j; i > 0 and j > 0, then an allocation of a slot at node Xði; jÞ for reception changes its state to Xði  1; j  1Þ. Such allocations occur with the rate ðki;j =2Þ. If i < j and j > 0, then an allocation of a slot at node Xði; jÞ for reception changes its state to Xði; j  1Þ. Such allocations occur with the rate ðki;j =2Þ. An allocation at node Xði; jÞ not only changes the state of X but also affects the states of nodes which are in its interference range. For example, suppose X 1 is a one-hop neighbor of node X and its state is ði1 ; j1 Þ. If i1 > 0 and j1 > 0, then an allocation of a slot for transmission at node X changes the state of X 1 to ði1  1; j1  1Þ. Such allocations at interfering nodes occur with the rate ðkIi ;j =2Þ since during that 1 1

slot X 1 is neither able to receive nor transmit to any node successfully. Similarly, if i1 > 0, then an allocation of a slot for reception at node X changes the state of X 1 to ði1  1; j1 Þ. Such interfering allocations occur with the rate kIi ;j =2 for 1 1

the similar reason given for the previous case. If i < M and j < M, then deallocation of a slot which is reserved for transmission at node Xði; jÞ changes its state to Xði þ 1; j þ 1Þ. Such deallocations occur with the rate li;j =4. It also changes the state of X 1 ði1 ; j1 Þ to X 1 ði1 þ 1; j1 þ 1Þ (if i1 < M and j1 < M) and such deallocations at the interfering nodes occur with the rate li1 ;j1 =4. If i ¼ j, then deallocation of a slot which is reserved for reception at node Xði; jÞ changes its state to Xði þ 1; j þ 1Þ or

to Xði; j þ 1Þ. Such deallocations occur with the rate li;j =8. Otherwise if i < j and j – M, then deallocation of a slot which is reserved for reception at node Xði; jÞ changes its state to Xði; j þ 1Þ. Such deallocations take place with the rate li;j =4. Such deallocations also change the state of X 1 ði1 ; j1 Þ to X 1 ði1 þ 1; j1 Þ with the rate li1 ;j1 =8 or let X 1 remain in state X 1 ði1 ; j1 Þ with the rate

li1 ;j1 =8, if i1 < j1

and j1 – M; move X 1 ði1 ; j1 Þ to state X 1 ði1 þ 1; j1 Þ with the rate li1 ;j1 =4, if i1 < j1 and j1 ¼ M; otherwise (if i1 ¼ j1 , i1 < M, and j1 < M) let X 1 to remain in state of X 1 ði1 ; j1 Þ with the rate li1 ;j1 =4. Since the cumulative call arrival rate at a node X is Poisson distributed with the mean kX ; ki;j ¼ kX , the call departure rate li;j from node X is equal to nl where n is the total number of calls currently under the service in the interference range of node X and l is the service rate of a call. The state transition diagrams of a node X when i ¼ j and i – j with possible state transitions and parameters are shown in Figs. 5 and 6. From the state diagrams shown in Figs. 5 and 6, we can write the following state equations for the developed multidimensional Markov model. For boundary cases: i ¼ 0 and j ¼ 0

ð3li;j =4ÞQ ði; jÞ  ðki;jþ1 =2ÞQ ði; j þ 1Þ  ðkiþ1;jþ1 þ kIiþ1;jþ1 =2ÞQ ði þ 1; j þ 1Þ ¼ 0 i ¼ 0 and j ¼ M

Fig. 5. State transitions of a node (i – j).

Please cite this article in press as: L. Rajya Lakshmi et al., PRIME: A partial path establishment based handover management technique for QoS support in WiMAX based wireless mesh networks, Comput. Netw. (2015), http://dx.doi.org/10.1016/j.comnet.2015.03.013

L. Rajya Lakshmi et al. / Computer Networks xxx (2015) xxx–xxx

11

Fig. 6. State transitions of a node (i ¼ j).

ðli;j =4 þ ki;j =2ÞQ ði; jÞ  ðkIiþ1;j =2ÞQ ði þ 1; jÞ  ðli;j1 =4ÞQ ði; j  1Þ ¼ 0 i ¼ 0 and j ¼ M  1 to 1

ð7li;j =8 þ ki;j =2ÞQ ði; jÞ  ðki;jþ1 =2ÞQ ði; j þ 1Þ  ðkiþ1;jþ1 =2 þ kIiþ1;jþ1 =2ÞQ ði þ 1; j þ 1Þ  ðkIiþ1;j =2ÞQ ði þ 1; jÞ  ðli;j1 =4ÞQ ði; j  1Þ ¼ 0

ðki;j þ kIi;j þ 7li;j =8ÞQ ði; jÞ  ðki;jþ1 =2ÞQ ði; j þ 1Þ  ðli1;j =8ÞQði  1; jÞ  ðli1;j1 =2ÞQði  1; j  1Þ  ðli;j1 =8ÞQði; j  1Þ  ðkIiþ1;j =2ÞQ ði þ 1; jÞ  ðkiþ1;jþ1 =2 þ kIiþ1;jþ1 =2ÞQ ði þ 1; j þ 1Þ ¼ 0 In addition, M X M X Q ði; jÞ ¼ 0 i¼0 j¼i

i¼j¼M

ðki;j þ kIi;j ÞQði; jÞ  ðli1;j =4ÞQ ði  1; jÞ  ð5li1;j1 =8ÞQ ði  1; j  1Þ ¼ 0

In total we have ððM þ 1Þ  ðM þ 2ÞÞ=2 equations. By solving the above mentioned state equations we can obtain the steady state probabilities of all possible states of a node.

i ¼ j ¼ M  1 to 1

ðki;j þ kIi;j þ 3li;j =4ÞQ ði; jÞ  ðki;jþ1 =2ÞQ ði; j þ 1Þ  ðli1;j =8ÞQ ði  1; jÞ  ð5li1;j1 =8ÞQ ði  1; j  1Þ  ðkiþ1;jþ1 þ kIiþ1;jþ1 =2ÞQði þ 1; j þ 1Þ ¼ 0 i ¼ M  1 to 1 and j ¼ M

ðki;j þ kIi;j þ li;j =4ÞQði; jÞ  ðkIiþ1;j =2ÞQ ði þ 1; jÞ  ðli1;j =4ÞQ ði  1; jÞ  ðli1;j1 =2ÞQði  1; j  1Þ  ðli;j1 =8ÞQ ði; j  1Þ ¼ 0 For non-boundary cases:

i ¼ M  2 to 1; j ¼ M  1 to 2 and i – j

6.1. Call dropping probability of a single hop call The steady state probabilities are used to find the theoretical bounds of the call dropping probability (CDP) of the handover calls. Using the steady state probabilities, first we compute the call dropping probability of a single hop call; then the call dropping probability of a multi-hop call is obtained using the call dropping probability of a single hop call. Suppose the partial path for a handover call is from X to its neighbor Y. A partial path with bandwidth requirement of k slots is to be established between X and Y. For the successful single hop path establishment between X and Y; k slots must be free for transmission at X and the same k slots must be free for reception at Y. Let us first obtain, the CDP of

Please cite this article in press as: L. Rajya Lakshmi et al., PRIME: A partial path establishment based handover management technique for QoS support in WiMAX based wireless mesh networks, Comput. Netw. (2015), http://dx.doi.org/10.1016/j.comnet.2015.03.013

12

L. Rajya Lakshmi et al. / Computer Networks xxx (2015) xxx–xxx

calls with bandwidth requirement of a single slot ðk ¼ 1Þ and later, the CDP of calls which require multiple slots. Partial path establishment between X and Y will fail if X is in any one of the states: ð0; MÞ; ð0; M  1Þ; . . . ; ð0; 0Þ (because no slot is available for transmission), or Y is in state ð0; 0Þ (because no slot is available for reception). So, the CDP of a call with bandwidth requirement of one slot is given by M X CDP ¼ ðQð0; iÞÞ:

ð2Þ

i¼0

From the CDP, the acceptance probability (CAP) of a call with bandwidth requirement of one slot between X and Y is given by

CAP ¼ 1 

M X

ðQ ð0; iÞÞ:

ð3Þ

i¼0

The probability of calls with bandwidth requirement of k slots instead of one slot to be dropped between X and Y is given by

! M k1 X X CDPðkÞ ¼ ðQ ðj; iÞÞ : i¼0

ð4Þ

j¼0

And the call acceptance probability is given by

! M k1 X X CAPðkÞ ¼ 1  ðQ ðj; iÞÞ : i¼0

ð5Þ

j¼0

6.2. Call dropping probability of multi-hop calls Assume that the partial path required for a handover call is not of single hop but of multiple hops and the bandwidth requirement of that call is one slot. To establish a multihop path between the nodes X and Y, the required bandwidth must be available along all hops of the path between X and Y [4]. To establish a partial path consisting of nodes X; X 1 ; X 2 ; . . . ; X N2 ; Y between X and Y, one slot must be available between the following pairs of nodes: ðX; X 1 Þ; ðX 1 ; X 2 Þ; . . . ; ðX N2 ; YÞ. For a handover call, to establish a partial path of N hops which consists of the nodes X; X 1 ; X 2 ; . . . ; X N2 ; Y, the required bandwidth must be available along all hops. Then the probability of successful establishment of a partial path of N hops is defined as:

CAP ¼ PðBandwidth ðk slotsÞ is av ailable between X and X 1 Þ  PðBandwidth ðk slotsÞ is av ailable between X 1 and X 2 j path establishment is successful up to X 1 Þ     PðBandwidth ðk slotsÞ is av ailable between X N2 and Y j path establishment is successful up to X N2 Þ: We assume that the events ‘‘Bandwidth (k slots) is available between X i and X iþ1 ’’ and ‘‘path establishment is successful up to X i ’’ are independent of each other. This assumption relaxes the upper bound on the call dropping

probability, since it causes over estimation of the number of free slots at a given node [4]. If the probability of successful establishment of a partial path of single hop is @, then from the above assumption the probability of successful establishment of a partial path of N hops is @ N . Then the failure probability of establishment of a partial path of N hops is given by

CDP  ð1  @ N Þ:

ð6Þ

6.3. Upper and lower bounds on CDP To derive the upper bound on the call dropping probability at a given call arrival rate, the steady state equations of the node, say node X, with maximum k=l (where k ¼ kX þ kIX ) and the maximum interference (that is, kIX ) are solved for steady state probabilities and used in the above equations to compute the upper bound on the CDP. The bandwidth available for handover calls at node X is less than the bandwidth available at the other nodes. As a result, such nodes drop more number of handover calls whose new paths are passing through them when compared with the other nodes. Similarly, the steady state equations of the node, say node Y, with minimum ðk=lÞ and the minimum interference are solved for steady state probabilities and are used to compute the lower bound on the call dropping probability. The bandwidth available for handover calls at node Y is more than the bandwidth available at the other nodes. As a result, such nodes drop less number of handover calls whose new paths are passing through them when compared with the other nodes. 7. Simulations and results The two metrics used to evaluate the performance of the proposed method are: partial path establishment delay and call dropping probability. The call dropping probability is defined as the fraction of handoff calls dropped due to the unavailability of QoS providing paths at the time of handoff. The partial path establishment delay is the time taken to establish a partial path with the required bandwidth from some crossover point to T-BS. We compare the call dropping probabilities of the proposed PRIME method and RFP based handover algorithm under the following conditions: ðiÞ while increasing the bandwidth requirements of the nodes gradually, ðiiÞ while increasing the call arrival rates of the nodes and ðiiiÞ while increasing the number of control slots available in a frame. In another set of simulations, the partial path establishment delays of PRIME are evaluated ðiÞ as the bandwidth requirements of the nodes and ðiiÞ as the number of control slots available in a frame increase. After the execution of PPEA, if a partial path is established to a moving MN, then PRIME provides lossless and seamless service to that MN. To prove this, two sets of simulations are conducted; in the first set the packet delay/packet drop rate during the handover is studied; in the second set the handover delay is computed to test whether it is within the required limits or not. However both the measures can be obtained in a single simulation set. In

Please cite this article in press as: L. Rajya Lakshmi et al., PRIME: A partial path establishment based handover management technique for QoS support in WiMAX based wireless mesh networks, Comput. Netw. (2015), http://dx.doi.org/10.1016/j.comnet.2015.03.013

L. Rajya Lakshmi et al. / Computer Networks xxx (2015) xxx–xxx Table 2 Simulation parameters.

Table 3 List of abrreviations.

Simulation parameter

Value

Bandwidth Frame duration Holdoff exponent Data modulation Bandwidth for VoIP calls Average bandwidth for VoD

10 MHz 4 ms 0 BPSK-1/2 12.8 kbps 64 kbps

addition, a set of simulations are conducted to compare the performance of the proposed single path establishment strategy and multi-path establishment strategy of PRIME in terms of the call dropping probability. Finally, the CPDs with simulations are compared with the analytical bounds. A detailed discussion of these simulation experiments is presented below. 7.1. Experimental setup and assumptions The ns-2 simulator (version 2.33) [16] is used for the validation of PRIME method. A patch for distributed scheduling of WiMAX backbone mesh networks is developed in [6] and is available as open source.2 That patch is used to implement the proposed architecture, PRIME method and RFP based handover method. All the buffers used for the algorithm implementation are assumed to be of infinite size. The simulation experiments are conducted on two different topologies (Random topology (RT) and Grid Topology (GT)) to compare the performance of PRIME and RFPHMT. In the first topology, nodes are placed at random positions, where as in the second topology nodes are connected in a grid like topology. In the random topology, 15 nodes are taken as BSs and 31 nodes are taken as clients, where as in the grid topology, 16 nodes are taken as BSs and 30 nodes are taken as clients. In both topologies nodes are placed in a square field of 2000 m  2000 m and the transmission range of a BS is set to 500 m. VoD calls are randomly generated between mobile nodes in an exponentially distributed call arrival rate. The average holding times of VoD services are exponentially distributed. VoIP calls of 100 s (exponential average) duration are generated by each mesh client in an exponentially distributed manner. Each experiment runs for 1500 s and all the presented results are averages of 50 runs. Some important simulation parameters are given in Table 2. The holdoff exponent parameter is used by the mesh election algorithm [8] to compute the transmission opportunities of nodes in the control subframes. 7.2. Performance comparison of PRIME and RFPHMT In this set of simulations, the CDPs of two handoff algorithms are compared while changing the various simulation parameters of both the topologies (RT and GT). The results of these experiments are discussed in detail below. One general observation from the results is that the CDP of RFPHMT is always more when compared with 2

13

http://cng1.iet.unipi.it/wiki/index.php/Ns2mesh80216.

PRIME WMN QoS MN BS MR TDMA RFP PMP LAN PBS FINCH CDG iCAR TWiLL IE MSH-DSCH PPEA NH PHOQREP PHOQREQ SPB LPI RFPHMT CDP CAP RT/GT MES PPE S-MN/D-MN S-BS/D-BS T-BS MNIH

Partial path establishment based handover Management Technique Wireless Mesh Network Quality of Service Mobile Node Base Station Mesh Router Time Division Multiple Access Race Free routing Protocol Point-to-Multi-Point Local Area Network Portable Base Station Fast Intra-Network and Cross-Layer Handover Client Data Group Integrated Cellular and ad hoc Relaying Throughput enhanced Wireless in Local Loop Information Element Mesh Distributed Schedule Partial Path Establishment Algorithm Neighborhood Pre-Handover Reply Pre-Handover Request Stop Bisection Last Packet Information Race Free Protocol Based Handover Management Method Call Dropping Probability Call Acceptance Probability Random Topology/Grid Topology Multipath Establishment Strategy Partial Path Establishment Source Mobile Node/Destination Mobile Node Serving Base Station of S-MN/Serving Base Station of D-MN Target Base Station of Destination Mobile Node Mobile Node Information for Handover

the CDP of PRIME. We noticed this observation in the case of random topology as well as the grid topology. 7.2.1. Call bandwidth vs. call dropping probability In this simulation experiment, we studied the call dropping probabilities of the networks as the bandwidth requirements of calls increase. The bandwidth requirements of the calls are increased gradually from 13 kbps to 78 kbps. The call arrival rates of all nodes are 1/2000 (calls/milliseconds). The results of this simulation experiment are shown in Fig. 7. It is expected that the call dropping probability increases with the increasing bandwidth requirements of the calls/nodes, since as the bandwidth requirements of the nodes increase, less bandwidth would be available for the handoff calls. One can understand clearly from Fig. 7 that the call dropping probability increases with the increasing bandwidth requirements of calls/nodes in the case of both topologies for PRIME and RFPHMT. But the call dropping probabilities of PRIME are less than that of RFPHMT for both topologies. 7.2.2. Call bandwidth vs. call dropping probability (with and without multiple paths) As the calls arrive and leave the network we reach a situation in which it is not possible to find a single partial

Please cite this article in press as: L. Rajya Lakshmi et al., PRIME: A partial path establishment based handover management technique for QoS support in WiMAX based wireless mesh networks, Comput. Netw. (2015), http://dx.doi.org/10.1016/j.comnet.2015.03.013

14

L. Rajya Lakshmi et al. / Computer Networks xxx (2015) xxx–xxx

Fig. 10. Number of control slots vs. call dropping probability. Fig. 7. Bandwidth vs. call dropping probability.

the call dropping probability when single path establishment strategy is used alone.

Fig. 8. Bandwidth vs. call dropping probability (random topology).

Fig. 9. Call arrival rate vs. call dropping probability.

path from a crossover point to the target base station of an MN especially when the bandwidth requirement of the MN is more. In this situation the multi-path establishment algorithm proves useful. It uses 2 or more paths to fulfill the bandwidth requirement of the handing off node and reduces the call dropping probability. In order to prove this fact two sets of simulations are conducted. In the first set we have not used the multipath establishment strategy (MES) of PRIME in the partial path establishment process. In the second set, MES is included in the partial path establishment process. As one can observe from Fig. 8, the call dropping probability when the single path establishment strategy is used together with MES is less compared with

7.2.3. Call arrival rate vs. call dropping probability In this simulation experiment the bandwidth requirements of all calls are set as 39 kbps. The call arrival rates of the nodes are increased from 1/12000 to 1/2000 (calls/ milliseconds). At the call arrival rate of 1/12000, since calls arrive at a lower rate, more bandwidth would be available for handoff calls, and the CDP should be low when compared with the CDP at the call arrival rate of 1/2000. This hypothesis is proved in this simulation experiment as shown in Fig. 9. In the case of RFPHMT, bandwidth has to be reserved along the path from the source to the target base station. Due to this reason, the call dropping probability with RFPHMT is more when compared with PRIME. 7.2.4. Number of control slots vs. call dropping probability In this set of simulations, the number of control slots available in a frame is increased from 2 to 10. ð1Þ When 2 control slots are available, a node is able to exchange less control information with its neighbors. As a result the path establishment request messages and partial path establishment request messages wait for a longer time at each node. Because of these waiting times, some of the path establishment requests and partial path establishment requests get dropped and result in high call dropping probability. ð2Þ When 10 control slots are available, nodes are able to exchange more control information with their neighbors which results in a lower call dropping probability. These two points (ð1Þ and ð2Þ) are proved in this simulation. When the new path establishment has to start from the source base station which is the case with RFPHMT, the accumulated delay along the path from the source base station to the target base station causes more handoff calls to drop. Results are shown in Fig. 10. 7.2.5. Fraction of calls accepted in the two-hop neighborhood As explained in Section 4, in PRIME, whenever T-BS has to establish a partial path for a mobile node, it first searches its neighborhood table for a crossover node. If it succeeds, the control overhead and the delay of path establishment are reduced significantly. Otherwise T-BS tries to find a partial path by broadcasting a pre-handover request. This simulation experiment studies the fraction of partial paths established within the two-hop neighborhood as

Please cite this article in press as: L. Rajya Lakshmi et al., PRIME: A partial path establishment based handover management technique for QoS support in WiMAX based wireless mesh networks, Comput. Netw. (2015), http://dx.doi.org/10.1016/j.comnet.2015.03.013

L. Rajya Lakshmi et al. / Computer Networks xxx (2015) xxx–xxx

15

Fig. 11. Bandwidth vs. CDP (random topology).

Fig. 13. Bandwidth vs. PPE delay.

Fig. 12. Bandwidth vs. CDP (grid topology).

the bandwidth requirements of calls are increased from 13 kbps to 78 kbps. As it is shown in Figs. 11 and 12, this experiment reveals the fact that as the bandwidth requirements of the calls increase the fraction of the partial paths established in the 2-hop neighborhood decreases. This observation is proved in the case of both topologies. 7.2.6. Call bandwidth vs. partial path establishment delay of PRIME We expect the partial path establishment (PPE) delays to increase with the increase in the bandwidth requirements of the nodes. As the bandwidth requirements of the nodes increase, less bandwidth is available at various nodes and it takes more time to establish a partial path. To prove this fact PPE delays are measured as the bandwidth requirements of nodes are increased from 13 kbps to 78 kbps and the results are shown in Fig. 13. We have observed a little peak at the bandwidth requirement of 65 kbps. In that case, as longer partial paths (partial paths with more number of hops) were established based on the bandwidth availability, it took little longer to establish those paths. 7.2.7. Number of control slots vs. partial path establishment delay of PRIME For this simulation experiment, the number of control slots available in a frame is increased from 2 to 10. When less control slots are available, a node is able to exchange less control information with its neighbors and results in more PPE delay. The PPE delay should be comparatively less when more control slots are available, as the nodes

Fig. 14. Number of control slots vs. PPE delay.

would be able to exchange more path establishment requests and partial path establishment requests with their neighbors. This fact is proved in our simulations as shown in Fig. 14. The PPE delay decreases with the increase in the number of control slots; with the increase in the number of control slots, the partial path establishment requests hit the crossover nodes quickly to complete the partial path establishment quickly. 7.3. Performance comparison of PRIME with the related research In addition to comparing the performance of the proposed method with the performance of RFPHMT, its performance is compared with the performance of a method ([19]) proposed in the literature. For that, instead of PPEA, PRIME is made to use the rerouting method proposed in [19] for handover management and this version is called as ‘‘ATM-R’’. According to this rerouting method, the nearest BS that falls on the old path and has a route to the target BS is used as the crossover BS to handle the handover of an MN. Call dropping probabilities of PRIME and ATM-R are evaluated as the call arrival rate increases from 1 to 5 calls/s. Random topology is used for the performance evaluation. VoD flows with the bandwidth requirement of 256 kbps are considered for this set of simulation

Please cite this article in press as: L. Rajya Lakshmi et al., PRIME: A partial path establishment based handover management technique for QoS support in WiMAX based wireless mesh networks, Comput. Netw. (2015), http://dx.doi.org/10.1016/j.comnet.2015.03.013

16

L. Rajya Lakshmi et al. / Computer Networks xxx (2015) xxx–xxx

Fig. 17. Time vs. packet delay.

Fig. 15. Call arrival rate vs. call dropping probability.

Fig. 18. Number of calls vs. call dropping probability.

Fig. 16. Number of control slots vs. handover delay.

should follow a decreasing trend as the number of control slots are increased. As shown in Fig. 16, this argument is correct; when sufficient number of control slots are available the proposed method achieves seamless handover. The flat regions in Fig. 16 indicate that with a small increase in the number of control slots in a frame, the handover delay does not vary much.

experiments. Data modulation schemes of the links are not fixed for this set of simulations. Links can use any one of the available modulation profiles. The results of the simulation experiments are shown in Fig. 15. It is observed that, as the call arrival rate increases, the CDPs of both the methods increase. The CDP of the proposed method is better than ATM-R. The partial paths between various nodes are fixed in the case of ATM-R. Whenever an MN performs handover from D-BS to T-BS, if the required bandwidth is not available along the partial path, ATM-R drops the respective flows. In contrast, PRIME tests most of the available partial paths to find a partial path with the required bandwidth. Even though the required bandwidth is available along the other partial paths, ATM-R does not consider those paths and results in a higher CDP. In contrast, PRIME by considering all available partial paths, results in a lesser CDP when compared with ATM-R.

In this simulation experiment, the packet delivery delays of a flow of a mobile node during its handover are plotted. The call arrival rate is 1/10000 (calls/milliseconds) and the bandwidth requirement of the calls is 78 kbps. Handover is taken place between 8.71063 and 8.77988. No packets are dropped during the handover because of the buffering of packets at T-BS. As one can observe from Fig. 17, there is not much difference in the packet delivery delays before and after the handover since a path through T-BS is available for the mobile node by the time of its handover to T-BS. If PPEA succeeds in the partial path establishment, then loss-less service can be continued to the MNs after they perform handover.

7.4. Number of control slots vs. handover delay

7.6. Comparison of simulation and analytical results

In this simulation experiment, the handover delay is plotted as the number of control slots are increased from 3 to 14. When less number of control slots are available, the delay for the handover should be high due to less control traffic exchange between the nodes and the delay

Using the Markov model which is explained in Section 6 and the formula (6) for call dropping probability, the upper and lower bounds on the call dropping probability of the handover calls are obtained as the bandwidth requirements of the calls are increased from 13 kbps to 78 kbps

7.5. Time vs. packet delay

Please cite this article in press as: L. Rajya Lakshmi et al., PRIME: A partial path establishment based handover management technique for QoS support in WiMAX based wireless mesh networks, Comput. Netw. (2015), http://dx.doi.org/10.1016/j.comnet.2015.03.013

L. Rajya Lakshmi et al. / Computer Networks xxx (2015) xxx–xxx

17

Fig. 19. Number of calls vs. call dropping probability.

Fig. 23. Number of calls vs. call dropping probability.

Fig. 20. Number of calls vs. call dropping probability.

and the CDPs of 3 hop partial paths are plotted in Figs. 18– 23. The respective simulation results are also plotted in each figure. In each simulation experiment, the bandwidth requirements of the calls are the same, but the call arrival rates are increased gradually. As can be seen from the figures, the results obtained from the simulations are within the theoretical bounds obtained from the multi dimensional Markov model. As the bandwidth requirements of the calls increase, the theoretical bounds on call dropping probability increase. This can be observed from Figs. 18–23. The same behavior is observed in the simulation results and the call dropping probabilities of simulations are always within the theoretical bounds. 8. Conclusion

Fig. 21. Number of calls vs. call dropping probability.

Fig. 22. Number of calls vs. call dropping probability.

In this paper a handover management method for the WiMAX mesh networks is proposed to continue QoS-constrained flows to the mobile nodes during their handovers. Different from the previous works which handle handover management in WiMAX mesh networks, PRIME addresses handover management in TDMA WiMAX WMNs deployed with distributed scheduling. In order to compare the performance of PRIME, another method named RFPHMT, which does not use the concept of crossover nodes and in which source/destination establishes the new path(s) for a moving destination/source is devised. To analyze the proposed method, a multi-dimensional Markov model is developed and theoretical upper and lower bounds on the handover call dropping probability are obtained. Using extensive simulation experiments, performance of PRIME and RFPHMT is evaluated in terms of call dropping probabilities and call setup delays. It is proved that the performance of PRIME is always superior to the performance of RFPHMT. Also, the simulation results show that the call dropping probabilities when using the single path establishment strategy together with the multi-path establishment strategy are less when compared with the call dropping probabilities when using single path establishment strategy alone. Simulations related to the partial path establishment delay show that the new paths are established to the mobile nodes quickly if enough number of control slots are available. It is also observed that the call dropping probabilities obtained through simulations are always within the theoretical bounds.

Please cite this article in press as: L. Rajya Lakshmi et al., PRIME: A partial path establishment based handover management technique for QoS support in WiMAX based wireless mesh networks, Comput. Netw. (2015), http://dx.doi.org/10.1016/j.comnet.2015.03.013

18

L. Rajya Lakshmi et al. / Computer Networks xxx (2015) xxx–xxx

Acknowledgment The authors would link to thank Prof. Swades De for his comments and suggestions. The authors gratefully acknowledge the valuable comments and suggestions from the anonymous referees. References [1] I.F. Akyildiz, X. Wang, Wireless Mesh Networks, John Wiley & Sons, 2009. [2] Y. Amir, C. Danilov, M. Hilsdale, R. Musaloiu-Elefteri, N. Rivera, Fast handoff for seamless wireless mesh networks, in: Proceedings of the 4th ACM International Conference on Mobile Systems, Applications and Services (MobiSys ’06), 2006, pp. 83–95. [3] J.G. Andrews, A. Ghosh, R. Muhamed, Fundamentals of WiMAX: Understanding Broadband Wireless Networking, Pearson Education, Inc., New Jersy, 2007. [4] A. Antony Franklin, V. Bukkapatanam, C. Siva Ram Murthy, Impact of multiple channels and radios on the performance of a TDMA based wireless mesh network, Comput. Netw. 53 (12) (2009) 2208–2224. [5] A. Boukerche, Z. Zhang, A hybrid-routing based intra-domain mobility management scheme for wireless mesh networks, in: Proceedings of the 11th International Symposium on Modelling, Analysis and Simulation of Wireless and Mobile Systems (MSWim’08), Canada, 2008, pp. 268–275. [6] C. Cicconetti, I.F. Akyildiz, L. Lenzini, FEBA: a bandwidth allocation algorithm for service differentiation in IEEE 802.16 mesh networks, IEEE/ACM Trans. Netw. 17 (3) (2009) 884–897. [7] R. Huang, C. Zhang, Y. Fang, A mobility management scheme for wireless mesh networks, in: Proceedings of IEEE Global Telecommunications Conference (GLOBECOM ’07), November 2007, pp. 5092–5096. [8] IEEE Std. 802.16-2004, Part 16: air interface for fixed broadband wireless access systems, IEEE, October 2004, pp. 1–857. [9] IEEE Std. 802.16e-2005, Part 16: air interface for fixed and mobile broadband wireless access systems – amendment for physical and medium access control layers for combined fixed and mobile operation in licensed bands, IEEE, December 2005, pp. 1–822. [10] I. Jawhar, J. Wu, A race-free bandwidth reservation protocol for QoS routing in mobile ad hoc networks, in: Proceedings of the 37th Annual Hawaii International Conference on System Sciences (HICSS ’04), January 2004, pp. 1–10. [11] P. Li, X. Yi, Y. Pan, A seamless handover mechanism for IEEE 802.16e systems, in: Proceedings of International Conference on Communication Technology (ICCT ’06), November 2006, pp. 1-4. [12] V. Navda, A. Kashyap, S.R. Das, Design and evaluation of iMesh: an infrastructure-mode wireless mesh network, in: Proceedings of the 6th IEEE International Symposium on World of Wireless Mobile and Multimedia Networks (WoWMoM ’05), June 2005, pp. 164–170. [13] M. Ren, C. Liu, H. Zhao, T. Zhao, W. Yan, MEMO: an applied wireless mesh network with client support and mobility management, in: Proceedings of IEEE Global Telecommunications Conference (GLOBECOM ’07), November 2007, pp. 5075–5079. [14] H. Wang, Q. Huang, Y. Xia, Y. Wu, Y. Yuan, A network based local mobility management scheme for wireless mesh networks, in: Proceedings of IEEE Wireless Communications and Networking Conference (WCNC ’07), March 2007, pp. 3792–3797. [15] J.-H. Yeh, J.-C. Chen, P. Agrawal, Fast intra-network and cross-layer handover (FINCH) for WiMAX and mobile Internet, IEEE Trans. Mobile Comput. 8 (4) (2009) 558–574. [16] K. Fall, K. Varadhan, The ns-2 manual, The VINT Project, January 2009, pp. 1–430. [17] L. Kleinrock, Communication Nets: Stochastic Message Flow and Delay, Dover Publications Inc., 1972. [18] M. Veeraraghavan, M.J. Karol, K.Y. Eng, Mobility and connection management in a wireless ATM LAN, IEEE J. Select. Areas Commun. 15 (1) (1997) 50–68. [19] C.-K. Toh, A hybrid handover protocol for local area wireless ATM networks, Mobile Netw. Applications 1 (3) (1996) 313–334. [20] M. Kim, I.-K. Ra, J.-S. Yoo, D.-W. Kim, H.-S. Kim, A seamless layer-2 handover scheme for mobile WiMAX based wireless mesh networks, in: Proceedings of International Conference on Advanced Communication Technology (ICACT ’09), 2009, pp. 974–979. [21] W. Du, W. Jia, W. Lu, Backtracking based handoff rerouting algorithm for WiMAX mesh mode, in: Proceedigns of the Third International

[22]

[23]

[24] [25]

[26]

[27] [28] [29]

[30]

Conference on Ubiquitous Intelligence and Computing (UIC ’06), 2006, pp. 341–350. M. Kim, J.-M. Kim, H.-S. Kim, I.-K. Ra, A proxy mobile IP based layer-3 handover scheme for mobile WiMAX based wireless mesh networks, in: Proceedings of Second International Conference on Ubiquitous and Future Networks (ICUFN ’10), 2010, pp. 33–38. B.S. Manoj, V. Mythili Ranganath, C. Siva Rama Murthy, On using multidimensional Markov chains for performance evaluation of hybrid wireless networks, IEEE Trans. Parallel Distrib. Syst. 17 (12) (2006) 1496–1511. H. Wu, C. Qiao, Modeling iCAR via multi-dimensional Markov chains, Mobile Netw. Appl. 8 (3) (2003) 295–306. I.F. Akyildiz, J. Mcnair, J.S. Ho, H. Uzunalioglu, W. Wang, Mobility management in next-generation wireless systems, Proc. IEEE 87 (8) (1999) 1347–1384. T.-J. Tsai, J.-W. Chen, IEEE 802.11 MAC protool over wireless mesh networks: problems and perspectives, in: Proceedings of the 19th International Conference on Advanced Information Networking and Applications (AINA ’05), 2005, pp. 60–63. Tropos Networks, . Nortel Networks, . R. Jayaraman, G. Raja, R. Arul, R. Kothandaraman, S. Sinto, A survey on scheduling mechanisms in broadband wireless networks, Lect. Notes Inform. Theory 2 (3) (2014) 214–219. K.S. Vijayalayan, A. Harwood, S. Karunasekera, Distributed scheduling schemes for wireless mesh networks: a survey, ACM Comput. Surveys 46 (1) (2013). Article No. 14.

L. Rajya Lakshmi received B.Tech. degree in Computer Science and Engineering from National Institute of Technology Warangal, India and M.Tech. degree in Computer Science and Technology from Jawaharlal Nehru University New Delhi, India. She received Ph.D. degree in Computer Science and Engineering from Indian Institute of Technology Delhi. Her research interests are handoff management in multi-hop wireless networks, routing, scheduling and call admission control techniques in wireless networks.

Vinay J. Ribeiro received B.Tech. degree from Indian Institute of Technology Madras and M.S. degree and Ph.D. from Rice University, all in Electrical Engineering. He is currently an Associate Professor in the Department of Computer Science and Engineering, at the Indian Institute of Technology Delhi. He has performed research internships at AT&T Labs, Sprint ATL, and Institut Mittag–Leffler. His research interests are in wireless networks (Wi-Fi, WiMAX, ad hoc, sensor), network traffic modeling, network tomography and queuing theory.

B.N. Jain received B.Tech. degree from Indian Institute of Technology Kanpur, and Ph.D. from SUNY, Stony Brook (NY), both in Electrical Engineering. Currently he is the vice chancellor of Birla Institute of Technology and Science Pilani. His research interests are in higher-layer protocols and 0-configurability issues of ad hoc networks, sensor networks, and applications to disaster mitigation and management, high-speed networks (MPLS, path protection) and network security.

Please cite this article in press as: L. Rajya Lakshmi et al., PRIME: A partial path establishment based handover management technique for QoS support in WiMAX based wireless mesh networks, Comput. Netw. (2015), http://dx.doi.org/10.1016/j.comnet.2015.03.013