With recent performance advancements in computer and wireless communications technologies, advanced mobile wireless computing is expected to see increasingly widespread use and application, much of which will involve the use of the Internet Protocol (IP) suite. The vision of mobile ad hoc networking is to support robust and efficient operation in mobile wireless networks by incorporating routing functionality into mobile nodes. Such networks are envisioned to have dynamic, sometimes rapidlychanging, random, multi hop topologies which are likely composed of relatively bandwidth-constrained wireless links.

Within the Internet community, routing support for mobile hosts is presently being formulated as "mobile IP" technology. This is a technology to support nomadic host "roaming", where a roaming host may be connected through various means to the Internet other than its well known fixed-address domain space. The host may be directly physically connected to the fixed network on a foreign subnet, or be connected via a wireless link, dial-up line, etc. Supporting this form of host mobility (or nomad city) requires address management, protocol interoperability enhancements and the like, but core network functions such as hop-by-hop routing still presently rely upon pre-existing routing protocols operating within the fixed network. In contrast, the goal of mobile ad hoc networking is to extend mobility into the realm of autonomous, mobile, wireless domains, where a set of nodes--which may be combined routers and hosts-- themselves from the network routing infrastructure in an ad hoc fashion. In this paper, the authors have compared OLSR and MOLSR protocols. The target of this protocol is to achieve lower delay and packet drop ratio and higher throughput. The paper is organized as follow: the first part explains the literature review, the second part introduces OLSR algorithm, the third part proposes the MOLSR in detail, the fourth part is the evaluation part and the last is the conclusion.

Multipath routing fits the serious time-varying characteristic of the mobile wireless network, which means it could provide effective bandwidth, respond to congestion and busty traffic, And increase Delivery Reliability. Because the classic Ad hoc on-demand Distance Vector routing (AODV) or the Dynamic Source Routing (DSR) performs better than Destination-Sequence Spread Spectrum (DSDV) [1,2] most of multipath algorithms are On-Demand based. Despite the fact that On-Demand routing could set up path faster, it needs to flood the network with route request. Furthermore, the disjoint path problem [3] and feedback information increase the overhead [4-6]. Another problem is the routing rediscovery. Even two paths were setup successfully, if the major path was broken, the alternate one always could not fulfill the QoS requirement by the wireless channel changing, so the transmitter may have to start the whole route discovery process all over again. Compared to On- Demand routing, though proactive routing broadcasts control messages in the network, if designed properly, those control messages could be effectively utilized to update the route table with only small overhead. The OLSR achieves both goals at the same time. The core parts of OLSR include multipoint relay (MPR), Topology Control (TC) message and the routing calculation. Each node in network selects a set of nodes called MPR in its neighborhood, which retransmitted its packets and TC message. Based on OLSR, we propose a proactive MOLSR, which introduces the cross-layer concept as well as the node discovery algorithm employed to discover every node on the route to avoid disjoint path [7]. The working of multipath OLSR in the sense of packet forwarding, route recovery and loop detection is derived from here [8,9].

2.1 Two Tables

There are two tables in OLSR. One is the topology table which records the organization of the whole network. An entry in the topology table includes an address of destination (MPR Selector), address of a last-hop node to that destination (originator of the TC message) and the corresponding MPR Selector set sequence number. The other table is the routing table which allows the node to route the packets for other nodes in the network. An entry in the routing table consists of destination address, next-hop address and the estimated distance to destination. These tables are not independent as they are related by the route calculation.

Apparently, there is no link state parameter e.g. SNR, BER in the topology table and the routing table, so precise routing selection is hard to be made.

2.2 Routing Calculation

Routing calculation in OLSR is a hop accumulation process, from hop 1 to h. The fundamental part is the looking up on the topology table and the contrasting with the existing routing table. So during the whole routing discovery process, there is no route request message broadcasting in the network, which is why the routing overhead in OLSR is much less than that in other proactive algorithms.

Although in the original OLSR routing calculation no specific route selection parameter is defined, the hidden choosing criteria is actually the hop-count, since for each entry in the topology table, if its destination address has appeared in the routing table, which means the hop count of this potential route is greater than that of the existing one, this potential route would be discarded no matter whether it was better than the existing route. Because the routing calculation begins from hop 1, the route with the same sender and receiver but greater hop count is hard to be recorded in the routing table.

3.1 TC messages Broadcasting and Topology Table

The neighbor sensing and MPR selection in MLOSR are the same as in traditional OLSR. However, after the MPR is decided, the TC messages and topology table are different from OLSR. In OLSR, nodes are not aware of the link state of other nodes two hops away. Two parameters, SNR and Delay, are added into the TC message in MOLSR. It is important to mention that the SNR and Delay in the TC message are the link state between MPR selector and TC originator. Because the route construction in the network is based on the collection of the TC message, when needed TC message arrived, the state of the whole route could be predicted and the route selection becomes reasonable.

As the TC message changes, the topology table which stores the TC message also needs to change. An entry in the topology table becomes an address of destination (an MPR Selector), address of a last-hop node to that destination (originator of the TC message), the corresponding MPR Selector set sequence number and the SNR/Delay of the link between MPR selector and TC originator (the SNR/Delay of the link between one hop neighbor of every node is also added in the neighbor table).

3.2 Multipath Routing Table Calculation

A node under MOLSR maintains a routing table which stores at most two routes to every destination in the network. These two routes are the best two paths that lead to that destination at that moment, and the node could choose one route to transmit data as the major route. If the major route collapsed, the other alternate route could be used immediately without another route discovery, providing better QoS than the single route OLSR. The most attractive feature of the multipath under MOLSR is that unlike other reactive multipath algorithm, alternate route detecting packets are not needed in MOLSR. The routing calculation is not only capable of building two routes for one destination but also updating those existing multipath routes. An entry in the routing table consists of destination address, next-hop address, the estimated distance to destination and the SNR/Delay of the route. The routing table is based on the neighbor table and the topology table, so the table will be re-calculated when a change in the neighbor table or topology table is detected.

The following procedure may be executed to calculate the routing table:

Step 1: All the entries in the routing table are removed.

Step 2: The new entries are recorded in the table starting with one hop neighbors (h=1) as destination nodes. For each neighbor entry in the neighbor table, whose link status is not uni-directional, a new route entry is recorded in the routing table where destination and next-hop address are both set to address of the neighbor and distance is set to 1 and the corresponding SNR/Delay is also recorded.

Step 3: Then the new route entries for destination nodes h+1 hops away are recorded in the routing table. The following procedure is executed for each value of h, starting with h=1 and incrementing it by 1 each time. The execution will stop if no new entry is recorded in iteration.

• For each topology entry in topology table, if its destination address does not corresponds to destination address of any route in the routing table AND its last-hop address of a route entry with distance equal to h, then a new route entry is recorded in the routing table where:
• The destination is set to destination address in the topology table;
• The next-hop is set to next-hop of the route entry whose destination is equal to above-mentioned lasthop address;
• The distance is set to h+1

The SNR/DELAY is set to:

(1)

(SNRT/DelayT is the SNR/Delay recorded in the corresponding topology entry)

• For each topology entry in topology table, if its destination address corresponds to destination address of one route in the routing table, then the node discovery algorithm will be executed:
• If this potential new route and the existing route only share two nodes, the sending node and the destination node, then this route is recorded in the routing table as above. The SNR/Delay of these two available routes is the criteria to decide which one is the major route or the alternate route.
• If this potential new route and the existing route share more than two nodes, then the route with better SNR/Delay among the existing route and this potential route will be reserved as above in the routing table and the other one will be discarded.
• For each topology entry in topology table, if its destination address corresponds to destination address of two routes in the routing table, then the node discovery algorithm will be executed:
• Among the two existing routes and this potential route, the two routes with only two same nodes and better SNR/Delay will be reserved and the other one will be discarded. The SNR/Delay of these two available routes is the criteria to decide which one is the major route or the alternate route.

Step 4: After calculating the routing table, the topology table entries which are not used in calculating the routes may be removed for memory saving sake.

3.3 Node Discovery Algorithm

OLSR is special in proactive routing protocols because it requires node to store a topology table other than just a routing table. The topology table gives the whole information about the structure of the network, facilitating the transmitter to confirm every node on the route more than just a destination node and a next-hop node. Most of the existing multipath routing protocols employ extra packets to detect and calculate the disjoint alternate route, which increase the overhead of the service. However, the alternate route detection and disjoint path decision become so easy under node discovery algorithm.

The node discovery algorithm is designed to let the sender discover all the nodes on the route, so the disjoint path could be decided. This algorithm executes as below:

Step 1: For one route and its corresponding entry in routing table where destination is A, next-hop is B, distance is h and SNR/Delay=SNR1/Delay1. Searching the topology table to find the entry where destination is A, last hop is D, sequence number is N and SNR/Delay=SNRn/Delayn

Step 2: If there is an entry where destination is D, next-hop is B and distance is h-1 in the routing table. Then execute:

(2)

• If this entry fulfills SNR/Delay= SNR'/Delay', then D is the node on this route which is h-1 hop away from the sender, and this node is recorded. Then the whole process starts with D to find the h-2 away node. Finally all the nodes on this route could be discovered.
• If several entries in the routing table fulfill all the requirements above, then the algorithm has to be executed on these entries.

There must be at least one route fulfilling this algorithm, because it is based on existing route. And the efficiency of this algorithm seems to decrease a lot under the (2) condition. However, the influence is limited. First of all, the wireless environment varies dramatically to different users, so it is very difficult for two different routes having the same SNR and Delay at the same time. In addition, though it seems that the calculation would increase a lot by the splitting of the path, most of the routes are the branch paths (not fulfill the disjoint requirements) and have been discarded during the calculation. Even there is the slightest possibility that at most two routes are left after the calculation. The amount of the disjoint path calculation and route selection would just increase by one route.

The above mentioned protocols are implemented to form a Virtual Class Room (VCR) [5]. A VCR is one that can be immediately established, and whose members can be dynamically added or removed; the group structure of the members can be reorganized dynamically. Figure 1 illustrates such an idea. The ad hoc classroom can support urgent and timely learning activities, thus improving learning effectiveness. For example [6], a teacher may establish a virtual classroom from his residence, students located around can take the opportunity to form an ad hoc group to improve the teaching learning process at any time using IEEE802.11g WLAN. VCR based on ad hoc network has been constructed [6] as shown in Figure 1 Based on JDK 1.5 platform, we structure a network which consists of 16 nodes.

Figure 1. A scenario of VCR using MANET

The application allows the user to initiate a query session with the peer or to lesson handling. The lesson file can include multimedia data like image, audio and video. Whenever a student (who is source of the communication session) wants to discuss any topic with other students or with a teacher (who is the destination of the communication session), they can initiate a query session by selecting the destinations from the member list displayed. To support transfer any type of file, UUEncode (Unix to Unix Encode) is used to convert the binary contents into plain ASCII characters, which can be transmitted over the network. At the destination side UUDecode is used to get back the original contents of the file.Old OLSR and MOLSR are tested under different transmission rate and mobility conditions. In the simulation, the transmitting rate is 10Mbits/s. It is obvious in Figure 2 that in the route construction phase when the Hello and TC messages begin to broadcast in the network, the delay of OLSR and MOLSR are almost the same, and after the route is built because the MOLSR selects better route and reserve an alternate route, the delay gets improved. Also Figure 3 and Figure 4 prove the MOLSR performs better in packet delivery ratio and throughput for the better route selection. Especially in Figure 3 when the 35th slot is received, there is an obvious jitter in the packet delivery ration in OLSR caused by possible route failure. However, the same problem gets solved in MOLSR where an alternative has been reserved.

Figure 2. The Delay of OLSR and MOLSR

Figure 3. The Packet Delivery Ratio of OLSR and MOLSR

Figure 4. The Throughput of OLSR and MOLSR

In this paper the authors compared the performance of MANET routing protocols. OLSR and Multipath OLSR are compared based on the parameters of delay, packet delivery ratio and throughput by implementing these protocols in a Virtual Class Room. Old OLSR and MOLSR are tested under different transmission rate and mobility conditions. In the simulation, the transmitting rate is 10Mbits/s and the results shows that MOLSR achieves lower delay and higher packet delivery ratio, throughput than the classic OLSR.