Opportunistic Routing Protocol (ORP)
5.4 Simulation Environment
The proposed protocol in section 5.3 has been implemented in Java and interfaced with MADHOC [HOGI] simulation tool. A number of 15,000 iterations / seconds, simulations were run to study the various conditions of the protocol based on many parameters. These parameters are discussed as follows.
5.4.1 Mobility Model
In mobile networks, devices are usually carried by humans so their movement is necessarily based on human decisions and social behaviour. To capture this kind of behaviour, it is assumed that people carrying the devices may form groups or move individually in the simulation area. Since movement is driven by social relationships, the simulation area is divided into a grid of 5 x 5 in the experiments. Each host moves in the simulation area using a Random Way-Point mobility model (RWP) [BROC98] [BETT02], RWP model is a very popular and frequently used mobility model in evaluation routing protocols for MANETs and has been extensively used in evaluation of routing protocols presented in [BALD05] [MAHE08] [GUID07] [MUTH05]. It is a simple and straightforward stochastic model that describes the movement behaviour of a mobile network node in a two–dimensional system area as follows:
The initial positioning of the nodes is typically taken from a uniform distribution. The nodes are typically placed in a square or a circular (disc) area.
A node randomly chooses a destination point in the area and moves with constant speed to this point.
After waiting a certain pause time, it chooses a new destination, moves to this destination, and so on.
The pause time durations are independent and identically distributed random variables.
Traveller nodes are also introduced to study the impact of higher mobility. Traveller nodes move between groups and share content with members of those groups.
95
5.4.2 Simulation Parameters
It is assumed that each user is equipped with a laptop device or a Wi-Fi enabled PDA device. Each device has an Omni directional transmission range of 100m. There are 100 users in a 1000m x 1000m environment. This environment consists of a 5 x 5 grid where each square size is 200m x 200m. The node speed is generated using a uniform distribution with values ranging [1, 5] m/s. The speed of the traveller node is set to 10 m/s. User may pause for up to 2 minutes to look for a destination.
32 different interest profiles are defined. Each user in the network would have to select up to four distinct interests; these interest profiles are randomly generated for the experiments. If one of the interests for two or more users is common, then these users are likely to start a conversation and share their documents. Since in real-life scenarios, users have various types‘ of interests and different types of documents to share (text documents, images, videos and audios), it is impossible to predict human behaviour and to the authors knowledge very few models exist that predict human social behaviour. For the sake of brevity, a set of interest profile determines the type of documents a user is interested in. Five types of documents each with a size limitation of up to 1024KB are defined. Every document created in the simulation is saved in the host‘s repository as (host_number_filenumber.ext) e.g. 4_F5_1.txt i.e. host number 4 creates document 1 of type F5. Documents are created every 100 seconds in the simulation as long as enough space is available in the repository. Figure 5.4 shows a matrix of interest profiles and type of documents a user may have as an interest. As an example, if a node has A0, C4, A3, and D2; as interest profiles, then it must be interested in document types 1, 2 and 4.
The size of the repository is set to 10MB maximum. Hosts broadcast an announce message every 15 seconds, this delay is introduced because at pedestrian speeds 15 seconds is generally considered as an adequate time for MANETs [HYYT06]. A node announces four interests in its profile, any neighbour with at least one of the similar interests, sends invite to share documents. At a certain time if the repository is filled and no further documents can be stored, the node in question would remove the least recently used document to make space for a newer document. This approach would permit a node to get rid of documents which have not been recently requested.
96
5.4.3 Compared Protocols
Content based routing in multihop networks has recently been an interesting research area. Not many researchers have worked in the area of content sharing in opportunistic data storing and forwarding. Some protocols have been presented for opportunistic data forwarding [PELU06] and routing in DTNs [JAIN04]. However these protocols do not address opportunistic content based data forwarding in delay tolerant MANETs. Content Based Multicast protocol (CBM) [ZHOU00], is also an opportunistic routing protocol but it also has an epidemic routing mechanism that relies on broadcasting messages, therefore is unsuitable for comparison. Baldoni et. al. in [BALD05] present a structure-less content based routing in MANETs. The proposed protocol uses frequent broadcasts for message delivery based on a complex estimation of proximity of potential subscribing nodes. The frequent usage of broadcasting increases flooding in the networks and reduces the effectiveness of the approach. Work presented in [YONE04] use a similar approach and report an adaptive content based routing protocol. The ORP protocol is specifically developed for content based routing in DTN while exploiting store-carry-forward mechanism for content delivery. Autonomous Gossiping (A/G) algorithm presented in [DATT04] is similar to the proposed protocol since it takes opportunistic approach for data transmission; nevertheless it relies on transmissions with the neighbouring nodes only. The A/G algorithm utilizes the epidemic algorithm to spread data items selectively based on vulnerability of other nodes (multicasting), instead of treating all nodes homogeneously and flooding the network. The A/G algorithm is considered for comparison with ORP protocol using transmissions only to neighbouring nodes. Table 5.1 shows the parameters used for comparison between the two protocols.
Figure 5.5 (a) shows a comparison of A/G algorithm with the ORP protocol, comparing the percentage of documents (content) delivered when strictly 2 or more profiles are matched.
Document types 1 2 3 4 5 Int er es t pr ofi les A0A1 A3A4 A6A7 B1B2 B4B5 C0 A5 B0 B3 B6 C1 C3 C6 D1 B7 A2 C4 C7 D2 D4 C2 C5 D0 D3 D5 D7 D6 D6 D7 D6
97 As can be seen, A/G performs better because it utilizes selective broadcast and multicast in propagating messages over the network. ORP considers multicast messages to neighbouring users (existing in n_list) only if more than 50% users have requested a document. Over a period of time the accuracy of documents received by ORP is better compared to A/G algorithm. However as shown in Figure 5.5(b) A/G creates far more number of documents compared to ORP, and therefore floods the network. This proves that although ORP is slower compared to A/G but is more effective due to selective multi- casting capabilities. Another point to be noted is for a secure application where trust management is of high importance, a selective multi-casting based protocol would provide better privacy for applications in delivering messages compared to a broadcasting protocol.
The A/G algorithm broadcasts/multicasts documents available in the storage area to requesting users; regardless of the significance of document to the receiver. This allows unsolicited documents to be sent without any request thus creating spam, increasing the amount of traffic in the network. Comparatively ORP protocol allows users to browse documents before being sent. Only requested documents are sent, therefore minimizing the amount of traffic generated. Further results for the comparison of ORP with A/G algorithm can be found in appendix A.