• Nenhum resultado encontrado

Friendship-based Routing Protocol for Delay Tolerant Networks

N/A
N/A
Protected

Academic year: 2023

Share "Friendship-based Routing Protocol for Delay Tolerant Networks "

Copied!
70
0
0

Texto

A dynamic threshold version was also proposed, where the friendship threshold changes over time as it corresponds to a portion of the best friend's weight. The results of this dynamic version were similar to those of the first version, only this time with about half the overhead.

Introduction

Context

The result was the creation of the Advanced Research Projects Agency (ARPA), now known as the Defense Advanced Research Projects Agency (DARPA). In the early 21st century, some of the ideas of IPN began to be applied to terrestrial communications and the term DTN came into use.

Document structure

Delay Tolerant Networks

  • Introduction
  • DTN Architecture Overview
  • Routing Protocols
    • Social-oblivious Protocols
    • Social-aware Protocols
  • Applications
  • The ONE Simulator

Traditional end-to-end routing algorithms are very efficient when there is a well-defined complete path from the source to the destination (eg the Internet). It is possible for multiple instances of the same bundle to coexist in the network, stored in the memory of multiple nodes. Socially aware protocols, on the other hand, rely heavily on the social relationships of nodes to route messages to the most promising next hop in terms of probability of delivery success.

As the name suggests, socially aware protocols examine the social behavior of the nodes that make up a DTN. Normally, a node is considered central if it plays an important role in the connectivity of the graph, that is, its proximity centrality is the sum of the shortest path distances between the node and all other nodes within it.

Betweenness centrality, on the other hand, takes into account the global structure of the network and can be defined as the number of shortest paths passing through a given node. The duration of this phase is denoted by 𝒕𝒂,𝒙, where 𝒙 is the number of the information transit segment. Naturally, most researchers in the field of DTNs tend to test protocols in this way.

Figure 2 – Differences between a DTN and TCP/IP network stack (extracted from [4]).
Figure 2 – Differences between a DTN and TCP/IP network stack (extracted from [4]).

Friendship Protocol

Concept

The friendship protocol is a social-based routing protocol for DTNs that exploits the ideas proposed in [15]. Since the source code adopted by the original authors was not available, the interpretation and implementation of this protocol is personal and adapted to take advantage of the ONE Simulator functionalities. The decision that a friendship is close or not comes down to the node that analyzes whether or not the friendship metric is above a certain threshold.

This protocol takes into account that node relationships often change over time periodically and addresses the fact that people's main activities often occur regularly, so friendship communities are periodic and respect a certain period or time of day.

The Protocol

Essentially, this algorithm 'corrects' the global time values ​​by defining a new origin of the time axis that corresponds to the beginning of the time slot in the matter. The next two algorithms, Algorithms 3 and 4, are part of the routines that are called when a connection is established or disconnected, respectively. In Algorithms 3 and 4, the variable totalS represents the current counter of the simplified SPM expression, obviously with respect to a particular time slot and a particular node.

If it was a node that has never been found before in the current time slot, the node initializes the time instant of the last encounter to a value of 0 seconds, as well as totalS with a value of zero, which means that the value of SPM is also 0 seconds. (4th and 5th lines). FNs always concern a specific time slot, resulting in nodes having 8 different FNs corresponding to 8 time slots of the day. In case the found node is the final destination of the message (line 2), the message is forwarded and deleted from the buffer.

If the destination of the message is part of the FN of the found node (line 7), then the message is sent and deleted from the buffer only if their friendship is stronger (line 8).

A dynamic threshold version

35 In this thesis, a modified version of friendship is also proposed, which assumes that the threshold for each node should be different. The idea is that the friendship threshold should instead be a fraction of its best friend's weight, rather than a fixed value that each node should accept. By comparing each friendship with its best friendship, nodes could adjust the number of nodes in their FNs according to the amount of activity (in other words, "nodes encountered") that each of them has.

Because the friendship weights of the nodes' friends change over time, the threshold would vary along, resulting in a dynamic that would follow each node's vision with respect to friends. To distinguish each version of the Friendship Protocol, the one proposed in this section is hereby referred to as Dynamic Friendship, as the other remains as the Classic Friendship version. To summarize, the only difference between these two versions is that in Dynamic Friendship the threshold is defined as a fraction of its best friend's weight, which must therefore be updated regularly.

More specifically, the threshold update in Dynamic Friendship occurs when there is a change in the connection status.

Simulation Results

Performance evaluation

In this thesis, the parameters to be evaluated are the delivery rate, the average delay of each message delivered and the overhead produced. Although achieving a decent delivery rate is the ultimate goal of any protocol of this nature, it is still interesting to consider the message delivery delay to find out how long it takes for a message to be delivered. The protocols chosen to be tested alongside Friendship are the Epidemic, Prophet and BubbleRap protocols, all of which have particular characteristics that may be disadvantages compared to Friendship.

The author of the source code used to implement the BubbleRap protocol is PJ Dillon [44] from the University of Pittsburgh, while for Epidemic and Prophet the source code was available natively in the ONE simulator. In the initial phase, the primary goal is to tune BubbleRap's parameters k and the Familiar Threshold and Friendship's Friendship Threshold to optimize their performance under the given scenario conditions. The Known Threshold in BubbleRap corresponds to the minimum total time a node must stay in contact with others to be considered part of its community, while the Friendship Threshold in the Friendship Protocol is the minimum friendship weight a node must achieve. to be considered a friend.

The second phase of the simulations is aimed at evaluating the effect of network traffic load by changing the number of created messages per time unit, now that we have optimized protocols.

Phase I: Tuning

  • Settings
  • Epidemic Protocol
  • Prophet Protocol
  • BubbleRap Protocol
    • Influence of the familiar threshold parameter
    • Influence of the k parameter
  • Friendship Protocol
    • Classic Friendship: Influence of the fixed threshold
    • Dynamic Friendship: Influence of the dynamic threshold

To determine the influence of the known threshold parameter, k was fixed at 4 and the known threshold was varied from 100 to 10000 seconds. It was observed that, with respect to the delivery rate, the best results occurred for higher values ​​of the known threshold. The dynamic version of the Friendship Protocol assumes the idea that each node should have its own friendship threshold level as part of its best friendship.

By reducing the fraction, communities in turn become more limited as the threshold approaches the weight of the best friendship. In terms of delivery rate, it is interesting to see that the most restrictive friendship communities corresponded to the lower performance, around 33%. The maximum delivery rate was recorded for a dynamic threshold of 80% of the best friendship weight, with a value of 35.23%.

In terms of the expense ratio, this version of the Friendship Protocol followed the pattern observed in the classic version in which overall expenses decreased whenever communities were more selective, i.e.

Figure 17 – Metropolitan area of Helsinki, Finland, presented on the GUI mode of The ONE simulator
Figure 17 – Metropolitan area of Helsinki, Finland, presented on the GUI mode of The ONE simulator

Phase II: Traffic load variation

  • Delivery Rate
  • Overhead ratio
  • Average Delay

The results of the evolution of the delivery rate metric with the message generation rate in the network are shown in Figure 18. In terms of the delivery rate, the results show that there is no clear benefit of using the dynamic version of the Friendship Protocol in place of the classic version. In conclusion, both versions of the Friendship Protocol had positive performances in terms of delivery rate metrics, as no other tested protocol showed better results according to the confidence interval.

It is also notable that the overhead ratio tends to decrease with the increase in message generation rate for each protocol. Regarding the overhead ratio metric, the benefits of using Prophet to the detriment of Epidemic are clearly observed, as the overhead is significantly lower for Prophet at all times. These results reveal the main advantage of using the dynamic version of Friendship to the detriment of the classic version, as the overhead ratio of the dynamic version is approximately half that of the classic for all prices and thus proves to be more efficient.

From the generation rate of 0.5 to 2 messages/minute, which corresponds to the lowest network load tested, the dynamic version of the Friendship Protocol stands out, as it presented the highest average delay, reaching a maximum of 35635 .04 seconds for the rate of 0.5 messages / minute.

Figure 19 – Overhead ratio vs. message generation rate
Figure 19 – Overhead ratio vs. message generation rate

Conclusions

53 The main objective of this thesis was to evaluate and compare the performance of the Friendship Protocol, a social aware DTN routing protocol based primarily on the social concept of Friendship. In this chapter, it is possible to understand what is the main idea behind this protocol and how it was implemented, illustrated with the pseudocodes of the most important algorithms and a flowchart. As an alternative to a fixed threshold approach, a second version of this protocol is proposed in which the threshold is a fraction of the current best friendship, meaning that the friendship standards vary from node to node.

In terms of delivery speed, both versions of the Friendship Protocol had a positive performance, as no other tested protocol showed better results according to the confidence interval for each tested network load value. The results also revealed that the overhead ratio of the dynamic version is about half that of the classic. Finally, it could also be worthwhile to create a version of the Friendship Protocol with multiple copies, as this can result in significant performance gains.

Raghavendra, Spray and wait: a efficient routing-skema for intermitterende forbundne mobilnetværk, Proceedings of the 2005 ACM SIGCOMM Workshop on Delay-tolerant networking, s.

Imagem

Figure 2 – Differences between a DTN and TCP/IP network stack (extracted from [4]).
Figure 3 – Example of a graph scheme
Figure 5 – Example of a Binary Spray-and-Wait (extracted from [2]).
Figure 6 – Community structures in a social graph, each represented with a different colour (extracted  from [11])
+7

Referências

Documentos relacionados

71% Perceptions and behaviours Shopping around 16% Checking if provider is regulated 11% Checking provider’s reputation 13% Dissatisified 13% Satisified 87% Dissatisified 26%