Protocol reliability through miss hit occurrence

The miss hit effect occurs when the system finds a positive match for a particular request description but the data type parameters that satisfy the request does not exist in the network.

If a protocol’s performance oscillates greatly in the results obtained, the protocol is considered faulty.

To simulate the behavior in the present work, 300 nodes in the network where randomly selected and assigned a total of 6 data type parameters (input & output) for each node to be used in the compatibility match and perform the service discovery task. A variable named FRR (False Response Ratio) controls the ratio of invalid search results produced during the active service discovery phase of the protocol. The higher the ratio, the more difficult it is to correctly discover the services in the network.

Figure 17 shows the performance of the service discovery protocol when the FRR value is increased and how the system behaves to the characteristics of the network for two different sets of nodes with mobility intervals of 200 seconds and 100 seconds. As specified in the section 5.1 of the current chapter, the configurations of the environment gives an 𝐴𝐴𝐴𝐴𝐴𝐴 value greater or equal to 5% following the equation presented in 〔20〕.

This implies that at any given moment, a total of close to 5% of nodes are disconnected from the network and hence, the percentage of services discovered in the network cannot reach the total of 100%. The reasons for this is that the connectivity paths between a node starting a request and a node providing a suitable service do not exist or it was broken, also, the TTL value of a request might not be enough to reach a service that fulfills the requirements of the same, without the chance to reach all the nodes in the network, the service discovery is reduced, another reason is the possibility that a node that started the search or one that offered the service

53

suddenly changed locations or left the network, resulting in the invalid search request message and request response communication.

Figure 17: Invalid search response effect in service discovery protocol

The results obtained show the performance of the proposed protocol for two different mobility intervals, the percentage of services found in the network remain stable during the simulation, the totality of services in the network cannot be achieved with the current characteristics of the environment used and for the reasons mentioned before. When the FRR reaches 1, the amount of services compatible with the searches is zero because there are no services matching any search request started in the network.

Figure 18 shows the average time to find a compatibility match between services and search requests in the network. Through the use of the data type table in each nodes and the data classified through the ontology, the compatibility match and search mechanism part of the protocol should maintain the quality of searches consistent throughout the increase of FRR in the simulation. When the FRR value reaches 1 there are no services matching any of the active requests in the network, hence the action is not triggered and no time is wasted in search process.

54

Figure 18: Invalid search response effect in compatibility match process time

Again the average time to find the first match of a service to a search request in the network remains stable with the increase of FRR value. The set of nodes with 200 seconds mobility interval perform better because of the extra time each node remains steady in the network, hence the communication path have less chance to be altered or be broken due to nodes moving or leaving the network.

Figure 19 shows the number of messages transferred in the network as a result of the message collisions and necessary re-transmission when the FRR value increases. Through the reliability in the dissemination layer explained in section 4.2.2 of the current chapter, the acknowledgement messages play an important role for the message transmission in the network, when the FRR increases the number of messages disseminated in the network decreases due to the absence of a positive response from the compatibility match of nodes in the network. When nodes receive a message, a response should be directed to the node that started the transmission, if the waiting time is higher than the set threshold, the entry for the expected response is erased, hence, when more invalid searches are placed in the network, the amount of search messages is decreased due to less possibility of message collision and the need of re-transmission of the same to invalid nodes in the network.

55

Figure 19: Invalid search response effect in number of search messages disseminated

Two different mobility intervals are used in the simulation and when the FRR equals to 1 there is no service compatible to any search request active in the network, hence the amount of search messages is zero. As shown in figure 18 the first match occurrence time remains constant although the number of search messages is decreased, as shown in figure 19, with higher values of FRR demonstrating the satisfactory performance of the protocol and the efficient adaptability to the dynamic characteristics of the network environment.