To ensure high delivery ratio and improve performance, widespread works have been carried out about buffer management in DTNs and opportunistic networks. This section briefly explains buffer management policies in the literature.
In the Drop Least Recently Received (DLR) policy, the packet that stays for a long time in the buffer will be dropped first [6]. In the Drop Oldest (DOA) policy, the message with the shortest remaining life time is dropped first. The idea behind this technique is, messages that are in the network from a long period of time have the high probability of having already been delivered [6]. In the Drop Largest (DLA) buffer management policy, the message with a large size will be dropped [6].
Drop Front (DF) buffer management technique drops the messages base on the order they enter into the buffer. For example, the first message that enters the buffer will be dropped first [7]. Evict Most Forwarded First (MOFO) tries to maximum diffusion of messages over the network by dropping those messages that have been sent the maximum number of times. This way, messages with less hop count are able to travel more within the network [7]. Evict Most Favorably Forwarded First (MOPR) maintains the value of each message in its queue. Each time when a message is replicated depending on the predictability of the message delivered, the message value will increase. So, the message with the highest value is dropped first [7]. Evict Shortest Life Time First (SHLI) buffer management technique uses the timeout value of the message, which indicates when it is no longer useful. This means that a message with the shortest remaining life time is dropped first [7]. Evict Least Probable First (LEPR) technique works by a node ranking the messages within its buffer on the basis of the predicted probability of delivery. The message with the lowest probability is dropped first [7].
In Threshold Drop (T-Drop) [8], if the buffer is full, a message within the calculated threshold will need to be dropped to make more free space for the new message.
In Flood based Drop (FBD) [9], The information collected during a message flood is used to decide whether to drop a packet.
In Least Recently Forwarded (LRF) [10], This policy releases a pre-sent packet that has been buffered for the longest time. This policy only applies to pre-sent packages. If none of the messages have been forwarded, then FIFO will be used.
In Global Knowledge Based Scheduling and Drop (GBSD) [11] Global information is used to calculate per-message utilities that help determine what needs to be dropped. The average delivery rate, and the average delay are both calculated. When calculating, the goal is to maximize the average delivery rate, and minimize the average delay. After calculating, the packet that has the smallest value is dropped.
In [12], authors investigated the impact of using social relationships to manage the buffering of OppNet nodes. First, they classified the relationships between the nodes in two groups: their friendship, and their acquaintanceship. Then, they proposed the Friendly Drop Algorithm (FDA), which in its decisions blends both friendship and acquaintanceship with the nodes. FDA explores the self-reported friendship relationships from the nodes to build friendship graphs. Moreover, FDA uses a metric based on the contact similarity among the nodes to find out their acquaintanceship.
In [13] the proposed algorithm has considered the probability of message delivery, buffer status, and message delivery time concurrently in selecting the relay node and in allocation the tokens sent to the node.
This [14] policy uses global information about the network either to maximize the average delivery rate or to minimize the average delivery delay. Then, an algorithm is created that uses statistical learning in order to estimate information about the global state of the network, and uses this estimation to approximate the optimal algorithm in practice.
In this article [15], authors integrated the different parts of buffer management, and take all information relevant to message delivery and the network resources into account. Based on statistics and analysis of the messages status, and considering the delivery history of the node and location information, combined with the relevant information from mutual learning between nodes, this article proposes a comprehensive integration buffer management strategy.
In [16], authors proposed a buffer management scheme that is inspired by the law of diminishing marginal utility in economics. This law shows that as a user increases consumption of a product, there is a decline in the marginal utility that user derives from consuming each additional unit of that product. They argued that generating excessive redundant copies in the network for a single message was not good, because this will occupy too much buffer space of nodes and will limit other messages opportunity to access the buffer space when buffer overflow happens. In contrast, in this proposed scheme, they estimated the state of a message, e.g., the total number of its copies in the network and its dissemination speed, and perform buffer replacement and scheduling accordingly. Messages that have larger estimated number of copies and faster dissemination speed are dropped prior to other messages when buffer overflow occurs and are forwarded posterior to other messages.
In [17] buffer drop strategy E-DROP is proposed to optimize the performance of DTN routing protocols in term of relayed, dropped, delivery probability, latency time averages, overhead ratio, hop count averages and buffer time averages.
In [18], buffer management policy is proposed for D2D multi-copy opportunistic routing algorithms named Space-Time Drop (ST-Drop). This policy checks local information to determine when and where a message is received over the network. The basic idea is that a message with a greater time and space coverage is more likely to have been delivered, so it can be dropped first.
In this paper [19], buffer management strategy is proposed by Merging the messages’ and nodes’ attributes, and migrate the redundant messages to the neighbor to optimize the total utility, instead of deleting them.
Le et al. [20] introduces a new DTN buffer management strategy based on the power distribution contacts law. The authors have focused on reducing message delay in delivering messages to the destination. They have considered two issues. The first issue is the order of repetition of the message that should be followed if the bandwidth is limited. The second issue is the selection of message to be dropped when buffer space is full. The authors have created a utility function to calculate average delay per packet. Scheduling of messages to be dropped is done on basis of the proposed utility function. The proposed scheme has reduced message latency and increased delivery ratio.
DF and DLA method maximize the drop but our proposed method decides whether or not to drop the message. MOFO maximizes the number of hops and increase latency, DLR has less probability of being conceded to the other nodes. In DOA method, messages are in the network from a long period of time. In some previous works, buffer management policy does not consider which message should be transmitted first or which node should be considered as sender. Our proposed buffer management policy improves or removes above mentioned drawbacks.