Reliable Low-Latency Multicasting in MANET: A DTN7-Driven Pub/Sub Framework Optimizing Delivery Rate and Throughput

Abstract

This paper addresses the challenges of multicasting in Mobile Ad Hoc Networks (MANETs), where communication relies exclusively on direct interactions between mobile nodes without the support of fixed infrastructure. In such networks, efficient information dissemination is critical, particularly in scenarios where an event detected by one node must be reliably communicated to a designated subset of nodes. The highly dynamic nature of MANET, characterized by frequent topology changes and unpredictable connectivity, poses significant challenges to stable and efficient multicasting. To address these issues, we adopt a Publish/Subscribe (Pub/Sub) model that utilizes brokers as intermediaries for information dissemination. However, ensuring the robustness of broker-based multicasting in a highly mobile environment requires novel strategies to mitigate the effects of frequent disconnections and mobility-induced disruptions. To this end, we propose a framework based on three key principles: (1) leveraging the Disruption-Tolerant Networking Implementations of the Bundle Protocol 7 (DTN7) at the network layer to sustain message delivery even in the presence of intermittent connectivity and high node mobility; (2) dynamically generating broker replicas to ensure that broker functionality persists despite sudden node failures or disconnections; and (3) enabling brokers and their replicas to periodically broadcast advertisement packets to maintain communication paths and facilitate efficient data forwarding, drawing inspiration from Named Data Networking (NDN) techniques. To evaluate the effectiveness of our approach, we conduct extensive simulations using ns-3, examining its impact on message delivery reliability, latency, and overall network throughput. The results demonstrate that our method significantly reduces message delivery delays while improving delivery rates, particularly in high-mobility scenarios. Additionally, the integration of DTN7 at the bundle layer proves effective in mitigating performance degradation in environments where nodes frequently change their positions. Our findings highlight the potential of our approach in enhancing the resilience and efficiency of broker-assisted multicasting in MANET, making it a promising solution for real-world applications such as disaster response, military operations, and decentralized IoT networks.

1. Introduction

Mobile Ad Hoc Networks (MANETs) are self-configuring, dynamic wireless networks that operate without relying on fixed infrastructure. Owing to their flexibility and adaptability, MANETs have found widespread applications in various domains, including disaster recovery [1,2], battlefield communications [3], and mobile networking [4,5]. However, the inherent characteristics of MANETs, such as node mobility and link instability, impose significant challenges on the design of efficient communication protocols [6,7,8].

Among the different communication paradigms in MANET, multicast communication is particularly crucial. Multicasting enables efficient data transmission to multiple recipients, reducing bandwidth consumption compared to multiple unicast transmissions. Although multicast protocols are well-established in wired and infrastructure-based wireless networks, adapting them to the highly dynamic MANET environment remains a complex challenge.

Our Contribution

In this study, we investigate publish/subscribe (Pub/Sub) multicast in MANET, where the sets of publishers and subscribers can be dynamically reconfigured. Our objective is to develop an efficient Pub/Sub mechanism that enhances resilience to node mobility, a fundamental challenge in MANET. To achieve this, we integrate concepts from Named Data Networking (NDN) and Delay-Tolerant Networking (DTN)—two established technologies that address content-based communication and intermittent connectivity, respectively.

NDN shifts the communication paradigm from “where to send” to “what to send” by utilizing content-centric routing instead of traditional IP-based addressing. This model naturally supports multicast communication, as content requests (interest packets) can be aggregated and satisfied by multiple sources through caching mechanisms. The inherent data and interest caching in NDN aligns well with content-based message dissemination, making it a suitable foundation for Pub/Sub multicast in MANET.

In contrast, DTN addresses the challenge of intermittent connectivity by employing the store-carry-and-forward paradigm, allowing nodes to buffer and transport messages across disconnected network partitions. We focus on DTN7 (https://dtn7.github.io/, accessed on 1 May 2025), a DTN protocol designed for highly fragmented networks. By facilitating message forwarding across temporarily disconnected network segments, DTN7 improves overall network connectivity and ensures reliable message delivery despite mobility-induced link disruptions.

Our approach improves resilience to node mobility by introducing broker-based Pub/Sub multicast, where brokers facilitate message dissemination between publishers and subscribers. Unlike conventional broker-based methods, which suffer from performance degradation due to mobility, our system dynamically relocates the broker function within the network in response to node movement. This relocation can be achieved using one of the following approaches:

• Broker Replication—Creating multiple broker instances to maintain service continuity.
• Broker Migration—Transferring the broker role between physical nodes, similar to mobile agent-based systems.

Both approaches involve a trade-off between reducing message loss caused by mobility and minimizing the maintenance overhead associated with broker delegation.

A key challenge in broker relocation is informing publishers and subscribers of changes in broker placement. Our system employs different strategies based on notification target size:

• For Publishers: Notifications are delivered via unicast to minimize network overhead.
• For Subscribers: Notifications leverage NDN interest packet propagation, ensuring that only nodes expressing interest in specific content receive updates.
• For Large-Scale Notifications: When updates are required by a significant portion of the system, broadcast-based dissemination may be considered; however, targeted notifications generally improve efficiency.

By integrating NDN-based multicast, DTN-based delay-tolerant forwarding, and dynamic broker delegation, our proposed system enhances the robustness of Pub/Sub communication in MANET, effectively mitigating the negative impact of mobility on message dissemination.

The proposed method was evaluated through simulations conducted using ns-3.44 (https://www.nsnam.org/releases/ns-3-44/, accessed on 1 May 2025). The results demonstrate that the dynamic generation of broker replicas in the proposed method significantly reduces the average latency for subscribers to receive published messages while also improving throughput. Additionally, the use of DTN7 in the bundle layer mitigates performance degradation, even in scenarios where all nodes exhibit frequent mobility.

The remainder of this paper is organized as follows. Section 2 reviews multicast protocols in MANET and the fundamental strategies they employ. Section 3 provides an overview of NDN and DTN7, the core technologies underlying the proposed method. Section 4 details the proposed method. Section 5 describes the simulation setup used to evaluate the effectiveness of the proposed method and presents the obtained results. Section 6 discusses related work on Pub/Sub in MANET. Finally, Section 7 concludes the paper and outlines directions for future research.

2. Challenges in Multicast Communication for MANET

This section explores the fundamental challenges and recent advancements in multicast communication for MANET. Research on MANET-based multicasting gained momentum in the early 2000s, driven by the proliferation of mobile devices such as smartphones and feature phones. The introduction of Bluetooth technology and the establishment of the Wi-Fi Alliance in 1999 further accelerated the exploration of MANET architectures leveraging general-purpose handheld devices. Many of the early protocols developed during this period laid the groundwork for modern multicast strategies in MANET. To contextualize recent developments, we begin by reviewing key contributions from past research that have shaped the evolution of multicast communication in MANET.

2.1. Historical Overview of Multicast Research in MANET

Multicast communication in MANET has been extensively studied to address the inherent challenges of node mobility and dynamic topologies. One of the earliest approaches, proposed by Zhou and Singh [9], introduced the Content-Based Multicast (CBM) protocol, in which senders transmit information to a predefined geographic area instead of targeting specific recipients. This design allows users to retrieve relevant data as needed, reducing dependency on stable end-to-end connections.

Building upon this concept, Meier and Cahill [10] developed STEAM, an event-driven middleware that incorporates both sender- and receiver-side filtering. This framework influenced later technologies, including Bluetooth Mesh (https://www.bluetooth.com/learn-about-bluetooth/feature-enhancements/mesh/, accessed on 1 May 2025), by enabling decentralized content dissemination. Similarly, Baldoni et al. [11] proposed a content-based multicast mechanism that tracks node contacts, a technique reminiscent of NDN interest packet routing. They later extended this work to develop an adaptive Pub/Sub system over structured overlay networks [12], improving multicast efficiency in highly dynamic environments.

Other notable approaches include Costa and Picco’s [13] hybrid multicast model, which combines deterministic event routing with randomized link selection to enhance message delivery in dynamic MANET settings. Additionally, Datta et al. [14] introduced Autonomous Gossip (A/G), a lightweight protocol that eliminates the need for structured overlays but at the cost of reduced delivery guarantees.

A significant step toward integrating Pub/Sub semantics into MANET multicast was taken by Yoneki and Bacon [15]. Their approach utilized Bloom filters and a dynamically formed overlay mesh based on the On-Demand Multicast Routing Protocol (ODMRP) to enable efficient content distribution. By clustering nodes using K-means clustering based on subscription similarity, their method facilitated fine-grained message delivery while maintaining scalability.

2.2. Classification of Multicast Strategies in MANET

Based on prior research, three primary strategies have emerged for achieving efficient multicast communication in MANET, particularly in high-mobility environments:

• Geographic Relay-Based Multicasting: This strategy involves transmitting messages to a designated geographic region, from which receivers retrieve relevant information as needed. By decoupling sender–receiver coordination, this method reduces overhead and enhances scalability, making it well-suited for scenarios where recipient mobility is high.
• Network-Wide Broadcast with Local Filtering: In this approach, messages are broadcast across the network without prior knowledge of the receiver set. Each node independently filters received messages based on their relevance, ensuring that only pertinent data is processed. This strategy is particularly effective for event-driven communication in highly dynamic environments, where pre-established multicast groups may be impractical.
• Interest-Based Overlay Multicasting: Nodes with common interests form an overlay network to facilitate selective message forwarding. To mitigate disruptions caused by mobility-induced disconnections, random link augmentation techniques are applied, ensuring persistent overlay connectivity and reliable message propagation. This method balances efficiency and robustness, making it suitable for content-centric applications.

These strategies serve as the foundation for designing robust multicast protocols in MANET, offering trade-offs between efficiency, scalability, and adaptability to network dynamics. As MANET applications continue to evolve, optimizing these approaches remains a key research challenge in ensuring reliable and efficient multicast communication.

2.3. Recent Advances in QoS-Aware Multicast Protocols

While the foundational strategies discussed above established the theoretical framework for MANET multicast communication, recent research has shifted focus toward Quality of Service (QoS) optimization and practical implementation challenges. The following studies represent significant advances in addressing these concerns through innovative protocol designs and optimization techniques.

Debnath et al. [16] investigated multicast communication in MANET with a focus on evaluating Quality of Service (QoS) metrics such as packet delivery ratio (PDR), delay, loss rate, overhead, and throughput. Their analysis, based on extensive NS-2 simulations under various network configurations, revealed that increasing the number of transmitting and receiving nodes results in higher overhead and delay, yet contributes positively to overall network throughput. Despite the inherent packet loss in MANET environments, their findings confirmed that multicast AODV consistently outperforms unicast protocols across diverse scenarios, delivering better QoS performance while maintaining lower overhead.

Building on the need for QoS-aware routing, Tran et al. [17] proposed a novel Deep Q-Network-based protocol called DQMR for multicast routing in Cognitive Radio Mobile Ad Hoc Network (CR-MANET). DQMR is designed to construct a shortest-path multicast tree with minimal end-to-end (E2E) cost while satisfying QoS constraints. The protocol guarantees high stability, low delay, reduced control overhead, and a high PDR. Notably, the authors derived a closed-form expression for E2E queuing delay under both the random waypoint and reference point group mobility models, and validated these expressions through simulation. Comparative results demonstrated that DQMR significantly outperforms multicast AODV in routing delay, control overhead, and PDR.

Addressing the challenge of cluster head selection in multicast routing, Suresh et al. [18] proposed an Optimal Route Selection (ORS) algorithm. This algorithm selects cluster heads and backup candidates, and computes optimal paths between cluster heads and member nodes based on trust coefficients and residual energy. Experimental evaluations showed that the ORS approach effectively enhances throughput and PDR while minimizing delay and jitter, outperforming existing methods. Recognizing the limitations of traditional MANET multicast routing in handling QoS constraints, Kumari et al. [19] introduced the HAPM algorithm, which combines Ant Colony Optimization (ACO) and Particle Swarm Optimization (PSO) techniques. By leveraging a dynamic queueing mechanism, HAPM improves route discovery and network performance. NS-2 simulation results showed that HAPM achieves superior QoS outcomes—including higher PDR, lower delay and overhead, fewer hops, and better throughput—compared to conventional approaches. Additionally, HAPM was demonstrated to maintain computational efficiency, confirming its suitability for practical deployment. To improve reliability and fault tolerance in MANET, Sivapriya et al. [20] developed the FFTRP protocol based on fuzzy logic. This method assesses node reliability and link robustness using fuzzy cost metrics to select optimal routes. Through simulations conducted with NS-2 (version 2.1b9) and MATLAB (version R2024a), FFTRP exhibited substantial improvements over traditional routing protocols, achieving reduced delay and overhead alongside a higher PDR. These recent advances demonstrate a clear evolution from the early theoretical foundations toward practical, QoS-oriented solutions. The integration of machine learning techniques, metaheuristic optimization, and intelligent decision-making systems represents a promising direction for future MANET multicast protocol development, addressing the increasing demands of modern mobile applications while maintaining the fundamental principles established by earlier research.

3. Foundational Technologies

3.1. Named Data Networking (NDN) Overview

Information-Centric Networking (ICN) has recently gained significant attention in both academia and industry. Originating from the Content-Centric Networking (CCN) paradigm introduced by Van Jacobson [21], ICN addresses the limitations of the current IP-based Internet, which is optimized for host-to-host communication rather than content retrieval, despite the latter constituting the majority of network traffic.

Among various ICN architectures, this paper focuses on Named Data Networking (NDN) [21]. NDN replaces IP addresses with hierarchical content names, enabling name-based routing and in-network caching. In NDN, nodes are interconnected through a network of NDN routers, and data retrieval follows the following steps:

• A data consumer (subscriber) generates an interest packet containing the name of the desired content and sends it into the network.
• NDN routers forward the interest packet towards the data producer (source node) using name-based routing.
• Upon reaching the producer, a data packet containing both the content name and the requested data is generated and signed by the producer.
• The data packet follows the reverse path of the interest packet back to the consumer.

To facilitate this process, each NDN router maintains three essential data structures:

• Pending Interest Table (PIT): Stores outstanding interest packets awaiting a response. Each entry records the requested data name along with the input and output interfaces.
• Forwarding Information Base (FIB): A routing table that maps content name prefixes to output interfaces. The FIB supports prefix-based routing and enables multiple forwarding options for each prefix.
• Content Store (CS): A temporary cache for data packets, reducing redundant transmissions. Cache replacement typically follows a Least Recently Used (LRU) strategy.

NDN routers forward packets based on the following principles:

• Interest packets propagate from downstream (consumer side) to upstream (producer side), while data packets follow the reverse path.
• Upon receiving an interest packet, the router first checks its CS. If cached data is available, it responds immediately. Otherwise, it checks the PIT. If a matching entry exists, the router records the incoming interface and discards the duplicate interest. If no match is found, the interest is forwarded upstream based on the FIB.
• To prevent redundant forwarding, an interest packet requesting the same content is forwarded only once, even if received from multiple downstream nodes.
• When a data packet arrives, the router looks up the corresponding PIT entry and forwards the data to all downstream interfaces listed in the PIT. The PIT entry is then removed, and the data is cached in the CS.
• Since large content objects often consist of multiple packets, interest packets function similarly to TCP acknowledgments (ACKs) in controlling data flow.

This architecture enables efficient content retrieval while mitigating key limitations of traditional IP-based communication.

3.2. DTN7 Overview

Routing protocols for Delay Tolerant Networks (DTNs) are designed to accommodate intermittent connectivity and dynamic topologies. In DTN, when a node receives a message, it attempts to forward it towards the destination along an available route. However, if the next-hop node is temporarily unreachable due to mobility or network disruptions, the message is stored at an intermediate relay node. This relay node stores the message until a viable communication link to the next-hop node becomes available, at which point forwarding resumes. This process continues until the message ultimately reaches its destination. This routing mechanism is known as store-carry-and-forward.

The fundamental unit of data transfer in DTN is a bundle, which functions similarly to a packet in traditional IP networks but is generally larger. The Bundle Protocol (BP) operates above the network layer in the protocol stack, managing bundle creation, transmission, and reception.

DTN7 is a protocol specifically designed for DTN implementations and exhibits the following key features:

• The Bundle Protocol (BP) in DTN7 defines the rules for bundle creation, transmission, and reception.
• DTN7 operates across diverse physical and network layers, enabling data transmission over various mediums such as Wi-Fi, Bluetooth, and satellite links.
• Whenever a node establishes a connection with a neighboring node, DTN7 evaluates whether any stored bundles should be forwarded. The routing is dynamic, with paths selected based on current network conditions.
• When large bundles are fragmented during transmission, DTN7 reassembles them upon receiving all fragments.
• Security mechanisms, including bundle encryption and node authentication, are incorporated to ensure data confidentiality and integrity.

In contrast to traditional MANET routing protocols, where a broken link immediately results in a route failure, DTN7 improves message delivery reliability. Since a node holding a bundle may move within the network, it increases the likelihood that a viable routing path to the destination will eventually be established, thereby enhancing overall message delivery success.

4. Proposed Method

The proposed method aims to achieve Pub/Sub multicast with high resilience to node mobility. Our approach is based on the observation that the impact of node mobility on performance varies depending on the role of the node. The protocol is designed according to the following key principles:

• Under conditions of moderate node mobility, we employ the Ad Hoc On-Demand Distance Vector (AODV) protocol at the network layer, combined with a dynamic broker migration mechanism. This mechanism ensures continuity of broker functionality even when connectivity with the original broker is lost, thereby enhancing the robustness of subscriber pull operations against mobility-induced disruptions.
• Upon broker migration, it is crucial to inform relevant nodes of the new broker location. We adopt differentiated notification strategies tailored to the specific roles of nodes within the network: For publishers, notifications regarding broker migration are delivered via unicast to minimize network overhead. Subscribers periodically broadcast interest packets to nearby nodes upon their own movement, thereby advertising their location. Migrated brokers utilize these interest packets to inform subscribers of the updated broker location effectively. In scenarios involving a large number of nodes requiring notification, or when subscribers have moved significantly beyond immediate reach, a large-scale flooding approach is utilized to disseminate the broker relocation information comprehensively.
• In highly dynamic environments characterized by extreme subscriber mobility, the network experiences substantial performance degradation under traditional protocols. To mitigate this, we replace the AODV protocol with DTN7 at the network layer. DTN7 employs a store–carry–forward mechanism, significantly improving resilience and maintaining acceptable message delivery rates despite severe disruptions caused by high node mobility.

An overview of the proposed method is illustrated in Figure 1. The key components involved in the proposed method are the broker, publisher, and subscriber, which are represented in green, yellow, and red, respectively. In addition, other nodes may participate in the system as message relays; these are shown in blue in the figure. The following subsections describe key innovations and critical considerations necessary to implement each primary policy effectively. To begin with, employing DTN7 at the network layer helps mitigate issues related to decreased message reachability. Nevertheless, this approach cannot ensure that messages will arrive at their intended destinations in the same order as they were initially dispatched. Such a constraint results from the protocol’s inability to control the physical movements of nodes responsible for carrying data bundles. Therefore, the proposed approach is incompatible with distributed algorithms predicated on the existence of FIFO-based message queues between processes.

4.1. Broker Replica Node Selection and Migration

This section introduces a detailed distributed replica node selection and migration algorithm specifically designed for NDN environments. The algorithm ensures optimal replica node selection and dynamic migration, maintaining data availability and reducing network overhead in environments with frequently changing topologies. See Figure 2 for an illustration.

4.1.1. Migration Trigger

Replica migration is primarily triggered by the detection of connectivity degradation between the broker and subscribers. When a broker node sends a response message, it anticipates an acknowledgment (ACK) within a short interval (typically 1 s). If the broker node fails to receive ACK consecutively for K messages, it records these as “unreachable” events. If such failures persist, indicating significant deterioration or disconnection of the communication link, the migration process is activated.

4.1.2. Reachability Filtering

Prior to initiating the node election, a reachability filtering mechanism is implemented to ensure candidates are within the same network partition as subscribers. Reachability is determined through

• Network discovery procedures.
• Verifying the existence of routes to the subscriber’s prefix in local routing tables.

Only nodes passing this reachability criterion are considered legitimate candidates for replica placement.

4.1.3. Node Scoring Methodology

Candidate nodes periodically broadcast their status, comprising node stability, request frequency, and resource utilization. Upon receiving these broadcasts, nodes calculate a comprehensive score based on the following formula:

$$\mathrm{Score}=\alpha S+\beta I-\gamma L$$

(1)

where:

• S (Stability): Reflects node stability calculated as
  $S=1-(\mathrm{number}\mathrm{of}\mathrm{neighbor}\mathrm{changes}/\mathrm{fixed}\mathrm{upper}\mathrm{limit})$.
• I (Interest): Measures local request density, calculated as
  $I=\left(\mathrm{number}\mathrm{of}\mathrm{local}\mathrm{requests}\right)/\left(\mathrm{highest}\mathrm{request}\mathrm{count}\mathrm{in}\mathrm{network}\right)$.
• L (Load): Represents node load, calculated as
  $L=\left(\mathrm{used}\mathrm{cache}\mathrm{space}\right)/\left(\mathrm{total}\mathrm{cache}\mathrm{space}\right)$.

The weights $\alpha ,\beta ,$ and $\gamma$—specifically configured as $\alpha =0.4,\beta =0.4,$ and $\gamma =0.2$—indicate the relative importance of stability, interest density, and load, respectively, ensuring that the selected node balances long-term stability, request responsiveness, and resource availability.

4.1.4. Election and Migration Procedure

The algorithm employs the following detailed steps for replica node selection and migration:

• Step 1: Status Broadcasting
  Each node broadcasts its status information (stability, request count, and load) periodically (e.g., every 10 s).
• Step 2: Score Calculation
  Upon receiving the broadcasts, each node independently calculates scores for all candidates based on the scoring formula provided.
• Step 3: Replica Election
  Since the scoring methodology is consistent across nodes, a consensus on the best node to host the replica is reached network-wide.
• Step 4: Migration Execution
  If the node selected through election differs from the current replica holder, the current node initiates migration by notifying the elected node. The elected node proactively requests the data, officially taking responsibility as the new replica host.

The detailed process of the Replica Selection and Migration Algorithm is shown in Algorithm 1, which depicts the complete decision-making workflow for dynamically relocating a broker replica when network conditions change.

Algorithm 1 Replica Selection and Migration (Mathematical Formulation)

Require: N (set of nodes), broadcastInterval, K Ensure: Optimal replica node   Initialization: Set $\mathrm{currentReplica}\leftarrow$ initial node   1: while true do   2:     /* 1. Status Broadcast */   3:     for each $n\in N$ do   4:         $statu{s}_n=(S_n,I_n,L_n)$        ▹ Broadcast stability, interest, load   5:     end for   6:     Wait for broadcastInterval   7:     /* 2. Candidate Selection */   8:     $C=\{n\in N|\mathrm{Reachable}(n,\mathrm{subscriber})=1\}$   9:     /* 3. Scoring */ 10:     for each $n\in C$ do 11:         ${\mathrm{Score}}_n=\alpha S_n+\beta I_n-\gamma L_n$ 12:     end for 13:     /* 4. Selection */ 14:     $n^{\ast }\leftarrow {argmax}_{n\in C}{\mathrm{Score}}_n$ 15:     /* 5. Migration Condition */ 16:     if $n^{\ast }\ne \mathrm{currentReplica}$ then 17:         NotifyMigration(currentReplica, $n^{\ast }$) 18:         $n^{\ast }$.RequestDataFrom(currentReplica) 19:         $\mathrm{currentReplica}\leftarrow n^{\ast }$ 20:     end if 21: end while 22: function Reachable(n, subscriber) 23:     if NetworkDiscovery(n, subscriber) ∧ RoutingTableCheck(n, subscriber.prefix) then 24:         return 1 25:     else 26:         return 0 27:     end if 28: end function 29: function NotifyMigration(oldNode, newNode) 30:     Send notification to oldNode and newNode 31: end function

This scoring and election method ensures the selected node balances stability, local demand, and resource availability effectively. Nodes with high scores optimally maintain replica availability and network performance, particularly suited to dynamic network environments such as NDN topologies.

4.2. NDN Interest-Based Broker Migration Notification

This section presents a detailed description of the proposed NDN interest-based mechanism designed to efficiently notify relevant nodes (publishers and subscribers) of the updated location of a broker node after migration events in MANET. See Figure 3 for an illustration.

4.2.1. Algorithm Overview

The algorithm addresses the critical issue of promptly updating the broker’s location information following its migration. The key innovation lies in exploiting the name-based routing and stateful forwarding characteristics inherent in NDN to reduce notification delays, minimize control overhead, and ensure reliable notification delivery. The notification strategy is tiered into three distinct mechanisms based on the recipient type and network conditions.

4.2.2. Notification Strategies

Once the migration of the broker node is completed, the new broker immediately notifies the publisher node through a unicast control message. This message contains essential identification information of the new broker, such as the node’s ID, name, or network address. Upon receiving this notification, the publisher instantly updates its routing table and starts directing subsequent data transmissions to the new broker location. This unicast mechanism leverages the uniqueness and known position of the publisher, ensuring timely and efficient notification delivery without unnecessary network overhead.

For notifying subscribers, the system employs a specialized interest packet mechanism intrinsic to the NDN architecture. Upon migration completion, the new broker generates a unique interest notification packet, typically named using either the prefix of the serviced content or a dedicated control prefix (e.g., /content/newbroker). This interest packet leverages existing Forwarding Information Base (FIB) entries and Pending Interest Table (PIT) states created during previous content requests by subscribers. This specially constructed interest notification automatically propagates along reverse paths previously established by subscriber requests, effectively reaching only nodes interested in the content. The subscribers receiving this interest immediately extract the new broker’s identification details (e.g., node ID or name) directly from the interest packet. Subsequently, subscribers update their local forwarding tables accordingly. This method is highly efficient because it requires no explicit knowledge of subscriber locations and leverages NDN’s stateful forwarding and name-based routing capabilities, significantly improving the precision and timeliness of notifications.

In exceptional cases characterized by the subscriber’s movement speed being too fast for the processing to keep up, widespread node distribution, or extensive network topology changes that impede timely interest packet propagation, the new broker may resort to a network-wide broadcast notification mechanism as a supplementary measure. Although this method incurs substantial network overhead due to its flooding nature, it ensures that all subscribers are informed promptly when standard interest notifications fail to cover all necessary nodes.

The detailed process of the BrokerMigrationNotification Algorithm is shown in Algorithm 2, which integrates the combined use of the aforementioned strategies, selecting the most suitable notification approach based on dynamic network conditions.

Algorithm 2 Broker Migration Notification (Mathematical Formulation)

Require: b (new broker), p (publisher), S (set of subscribers), G (network) Ensure: Notification delivery to p and S   Step 1: Unicast Notification to Publisher   1: if RouteExists(p) then   2:     $m_b\leftarrow ({\mathrm{ID}}_b,{\mathrm{Addr}}_b)$   3:     $\mathrm{SendUnicast}(p,m_b)$   4: end if Step 2: NDN Interest Notification to Subscribers   5: ${\mathrm{InterestName}}_b\leftarrow$ “/content/newbroker/” + IDb   6: $\mathrm{ForwardInterest}({\mathrm{InterestName}}_b,G)$ Step 3: Wait for Subscriber ACKs   7: $N_{\mathrm{informed}}\leftarrow \left|\{s\in S|{\mathrm{ACK}}_s\mathrm{received}\mathrm{within}{T}_{\mathrm{timeout}}\}\right|$ Step 4: Broadcast if Needed   8: if  $N_{\mathrm{informed}}<\left|S\right|$then   9:     $\mathrm{Flooding}(G,m_b)$ 10: end if

4.3. DTN7-Enhanced Broker Migration and Replica Selection Algorithm

This section introduces an advanced broker migration and replica selection strategy specifically designed for environments characterized by high subscriber mobility. The proposed solution leverages Delay-Tolerant Networking (DTN7) protocols to ensure robust and reliable content delivery under highly dynamic network conditions. See Figure 4 for an illustration.

4.3.1. Integration of DTN7 Protocol

The fundamental approach incorporates DTN7 nodes as custodians within the network. These custodian nodes are responsible for reliably storing (store) and subsequently forwarding (carry–forward) bundles when encountering subscribers or other custodian nodes. The forwarding procedure employs Status Reports or Custody Acknowledgments to confirm successful transmissions, thereby significantly improving delivery assurance.

4.3.2. Trigger Conditions for Replica Migration

Replica migration under the DTN7-enhanced approach occurs under specific, stringent conditions:

• If a subscriber’s message reception delay surpasses three times the 90th percentile of historical delays.
• If consecutive custody timeouts are detected.

Upon encountering either condition, the existing broker identifies the communication pathway as insufficient to guarantee timely delivery and initiates the replica selection and migration process.

4.3.3. Reachability Filtering and Node Evaluation

Candidate nodes within the subscriber’s connected sub-network are initially filtered using a Reachability Filter. Following successful filtering, nodes are scored using the established criteria of stability (S), interest (I), and resource load (L):

$$\mathrm{Score}=\alpha S+\beta I-\gamma L$$

(2)

The node receiving the highest score is subsequently selected as the new replica host.

4.3.4. Bundle Transfer via BPQ Range-Fetch

Following the election of a new replica node, bundles pending delivery to subscribers are batch-transferred using the Bundle Protocol Query (BPQ) range-fetch mechanism. This procedure ensures rapid and complete bundle migration from the old custodian to the newly designated custodian node, guaranteeing continuity of service.

4.3.5. Version Vector-Based Bundle Synchronization

After the migration and upon subsequent reconnection events, custodian nodes synchronize their bundle states using DTN7’s version vector mechanism. This synchronization effectively prevents duplicate transmissions and ensures consistency across replicas.

The detailed process of the described DTN7-enhanced broker migration strategy is shown in Algorithm 3:

Algorithm 3 DTN7-Enhanced Migration (Mathematical Formulation)

Require: b (current broker), S (set of subscribers), G (network) Ensure: Optimal replica $n^{\ast }$, completed migration   1: for each $s\in S$ do   2:     $d_s\leftarrow \mathrm{CurrentDelay}\left(s\right)$   3:     $d_s^{90}\leftarrow \mathrm{HistoricalDelay}90\mathrm{Percentile}\left(s\right)$   4:     if $d_s>3d_s^{90}$or$T_{\mathrm{custody}}\left(s\right)=1$ then   5:         /* Replica Election */   6:         $C\leftarrow$ set of nodes with reachability to s   7:         for each $n\in C$ do   8:            $S_n\leftarrow \mathrm{ComputeStability}\left(n\right)$   9:            $I_n\leftarrow \mathrm{ComputeInterest}\left(n\right)$ 10:            $L_n\leftarrow \mathrm{ComputeLoad}\left(n\right)$ 11:            ${\mathrm{Score}}_n=\alpha S_n+\beta I_n-\gamma L_n$ 12:         end for 13:         $n^{\ast }\leftarrow {argmax}_{n\in C}{\mathrm{Score}}_n$ 14:         /* Bundle Transfer */ 15:         $B_s\leftarrow$ set of pending bundles for s 16:         for each $b\in B_s$ do 17:            $\mathrm{SendBundle}(b,n^{\ast })$ and wait for ACK 18:         end for 19:         Update replica status for $n^{\ast }$ 20:         Maintain redundant cache at b 21:     end if 22: end for

The DTN7-enhanced replica migration algorithm significantly improves reliability and message delivery rates under high mobility scenarios. By explicitly addressing stringent network conditions with precise triggering mechanisms and integrating robust DTN bundle management strategies, the proposed method ensures consistent service quality even in highly fragmented or intermittently connected networks.

5. Evaluation

This section presents a thorough performance evaluation of the proposed Pub/Sub framework under three aspects. Section 5.1 evaluates the impact of the broker replica migration mechanism on multicast performance by comparing scenarios with and without dynamic broker relocation. Section 5.2 examines different migration notification strategies (traditional routing vs. NDN interest-based notifications) and their effectiveness in promptly informing network nodes of broker movements. Section 5.3 assesses the integration of the DTN7 protocol in high-mobility scenarios, contrasting it with a baseline MANET approach to demonstrate improvements in delivery reliability and throughput. We first describe the common simulation setup used in all experiments, and then detail each evaluation objective with its specific scenarios, results, and analysis.

Experimental Setup: All simulations were conducted in the ns3 network simulator using a mobile ad hoc network of 50 wireless nodes deployed in a 1 km × 1 km area. Nodes in the network are assigned distinct roles: one node functions as the Publisher, continuously generating content; another node serves as the Broker, initially holding a copy of the Publisher’s content and mediating its delivery; and a third node acts as the Subscriber, requesting and receiving the content. The other 47 nodes function exclusively as relay nodes, responsible for forwarding messages along multi-hop paths as needed. We assume that the Publisher and Broker nodes are fixed in place, and the Subscriber is the only mobile node. The Subscriber moves according to a random waypoint mobility model. At the start of each simulation, the Subscriber begins in proximity to the Broker, within the Broker’s direct wireless communication range. This ensures that initially the Subscriber can retrieve content directly from the Broker with minimal latency. The Publisher node continuously produces content at a rate of 50 messages per second, and the Subscriber issues a subscription (request) for the latest content every 2 s. Each request is served by the current Broker node. In scenarios with broker migration enabled, the system will dynamically hand off the broker role to a new node if the Subscriber moves far from the current Broker; in the baseline scenarios without migration, the original Broker remains static. We ran each simulation for a duration sufficient to allow the Subscriber to move out of the original Broker’s range and (when enabled) trigger at least one broker migration event. Throughout the simulation, all message send/receive events and timings were logged for post-analysis. We evaluated performance primarily in terms of end-to-end delivery delay and throughput, as these metrics reflect the latency and data delivery capacity of the multicast service. The Average Delay (in milliseconds) is measured from when the Subscriber sends a content request to when it receives the corresponding content, averaged over all requests. The Average Throughput (in messages per second) is the total number of content messages delivered to the Subscriber divided by the simulation time. An ideal Pub/Sub mechanism in this context should minimize delay while maximizing throughput.

5.1. Evaluation of Replica Migration Mechanism

In this experiment, we assess how the dynamic broker replica migration mechanism affects multicast performance, by comparing it against a static-broker baseline. We define two scenarios: Scenario A—without Replica Migration, and Scenario B—with Replica Migration. In Scenario A, the Broker remains fixed at its initial location near the Publisher for the entire simulation. As the Subscriber roams away, it must retrieve content from the original Broker via increasingly long multi-hop routes. This represents a traditional system with no proactive relocation of content. In Scenario B, the replica migration mechanism is enabled: when the Subscriber moves sufficiently far from the original Broker, the system elects a more suitable node (among the relay nodes) as a new Broker and transfers the content to that node. The selection is based on a composite stability score considering node mobility, local request density, and load. After this handoff, the Subscriber’s subsequent requests are served by the new Broker, which is topologically closer to the Subscriber. If the Subscriber continues moving, the mechanism can repeat, ensuring the Broker role always migrates toward the Subscriber’s vicinity. By comparing these two scenarios, we can quantify the benefits of broker migration in terms of latency and delivery rate.

5.1.1. Results and Analysis—With vs. Without Replica Migration

The results are compared between the two scenarios (with vs. without replica migration) to assess how the migration mechanism affects these aspects of performance.

We collected the average content retrieval delay and throughput in both scenarios to evaluate performance. Figure 5 shows that at a moderate Subscriber speed of 2.0 m/s, the average request–response delay drops from about 255 ms in the no-migration scenario to only 44 ms when the replica migration mechanism is enabled. Without migration (Scenario A), as the Subscriber travels farther from the fixed Broker, each request must traverse an ever-lengthening multi-hop path, incurring significant route discovery and queueing delays. In contrast, with migration (Scenario B), a new Broker is elected nearer to the Subscriber once the Subscriber moves out of range, which drastically shortens the communication path and thus restores low-latency delivery. The results confirm that broker replication/migration effectively curtails latency by keeping the content source close to the moving subscriber. Figure 6 compares the average throughput achieved with and without broker migration at the same subscriber speed (2.0 m/s). With the replica migration mechanism enabled, the system delivers about 49.12 msgs/s to the Subscriber, whereas the baseline without migration achieves only about 11.67 msgs/s. This corresponds to roughly a 4.2-fold increase in throughput when using dynamic broker migration. Two factors account for this substantial gain. First, by relocating the Broker closer to the Subscriber, the forwarding path is shorter on average, which reduces per-hop transmission loss probability and congestion. Second, the shorter path is less prone to route breakages, thereby avoiding frequent route repairs, retransmissions, and buffer overflows. As a result, in Scenario B nearly the full publisher rate (50 msgs/s) is successfully delivered to the Subscriber, whereas in Scenario A (static-broker) the effective delivery rate is throttled to about 12 msgs/s due to the long, frequently interrupted route. These findings demonstrate that the replica migration mechanism simultaneously lowers latency and boosts throughput under moderate mobility, by proactively moving the broker functionality closer to clients as network topology evolves.

5.1.2. Experiment of Subscriber Movement Speed on Message Delivery Rate with Replica Migration Mechanism

To evaluate the effectiveness of the replica migration mechanism across different subscriber speeds, experiments were conducted at subscriber velocities of 2.0 m/s, 2.4 m/s, 2.8 m/s, and 3.0 m/s, measuring both the message delivery rate.

Table 1. Impact of subscriber movement speed on message delivery rate with replica migration mechanism.

Subscriber Movement Speed (m/s)	Message Delivery Rate (%)
2.0	98.25
2.4	85.17
2.8	73.36
3.0	52.00

demonstrates the impact of subscriber movement speed on the message delivery rate under the replica migration mechanism. At the lowest tested speed (2.0 m/s), the system achieves 98.25 percent. However, as the subscriber speed increases, there is a notable decline in the delivery performance. At 2.4 m/s, the delivery rate falls to 85.17 percent, further decreasing to 73.36 percent at 2.8 m/s. Significantly, at 3.0 m/s, the delivery rate drops sharply to only 52 percent, indicating that the mechanism struggles to maintain reliable delivery at higher subscriber speeds.

The substantial performance deterioration observed at higher subscriber speeds (3.0 m/s and above) primarily results from the reduced duration of the contact window, defined as the time period during which the Subscriber remains connected to a particular broker node. Using the approximate relationship

$$T_{\mathrm{link}}={\displaystyle \frac{2R}{\pi v_{\mathrm{rel}}}}$$

(3)

In this formula, R specifically denotes the distance that the Subscriber must move before losing communication with the broker. Given that the Subscriber’s initial position is approximately $R/2=125\mathrm{m}$ from the Broker, this scenario corresponds to $R=125\mathrm{m}$ and a stationary Broker, and the Subscriber’s walking speeds $v=2.8$ and $3.0\mathrm{m}/\mathrm{s}$ yield relative speeds $v_{\mathrm{rel}}=v$; substituting gives $T_{\mathrm{link}}\left(2.8\right)=28.43\mathrm{s}$ and $T_{\mathrm{link}}\left(3.0\right)=26.53\mathrm{s}$. Link failure is declared only after $k=3$ consecutive retransmissions have timed out; with the measured RTT of 0.6 s, the retransmission interval is set to $\Delta t=1.2\mathrm{s}$, so link-loss detection consumes $T_{\mathrm{detect}}=k\Delta t=3.6\mathrm{s}$. Thereafter, each replica election round broadcasts the local $(S,I,L)$ metrics for 10 s, followed by 0.8–1.0 broadcast periods (8–10 s), for the distributed ranking to converge. During the preceding 22 s (detection + broadcast + convergence), the Publisher, sending at $50\mathrm{msg}/\mathrm{s}$ and 1 kB per message, accumulates $50\times 1\times 22\approx 1100\mathrm{kB}$ of backlog; including the queue’s residual data, the backlog is 600–1000 kB. ns-3 measurements show an application-layer throughput of ≈100 kB s⁻¹ (≈1 Mb s ⁻¹ at the PHY), so copying the backlog takes 6–10 s. Summing detection (3.6 s), broadcast (10 s), convergence (8–10 s), copying (6–10 s), and ≈2 s of scheduling overhead, the entire replica migration pipeline requires 30–35 s. At $2.8\mathrm{m}/\mathrm{s}$, the 28 s window is tight yet sporadically long enough to house the 30 s pipeline; when the speed rises to $3.0\mathrm{m}/\mathrm{s}$, the 26 s window becomes strictly shorter than the minimum 30 s pipeline, so copying is routinely interrupted mid-path. The Broker then restarts the k-count, triggering another detect–broadcast–converge–copy cycle and producing a “migration-latency avalanche.” Meanwhile, higher speed intensifies neighbour churn, depressing the stability metric

$$S=1-(\mathrm{number}\mathrm{of}\mathrm{neighbor}\mathrm{changes}/\mathrm{fixed}\mathrm{upper}\mathrm{limit})$$

(4)

and prolonged delays dilute the interest I; cache pressure inflates the load metric L; their oscillations reorder node scores every 8–10 s and can spawn transient “dual Brokers,” further wasting bandwidth. Consequently, the delivery rate plunges from $74\%$ at $2.8\mathrm{m}/\mathrm{s}$ to $52\%$ at $3.0\mathrm{m}/\mathrm{s}$. Thus, these results highlight a critical mobility threshold (around 2.8 m/s) beyond which the effectiveness of the current replica migration mechanism substantially diminishes due to insufficient contact duration for the migration process.

5.2. Evaluation of Efficient Broker Migration Notification

This experiment evaluates strategies for notifying system nodes about broker migration events. The objective is to compare the NDN interest-based notification mechanism proposed in the presented framework with a more conventional routing-based approach—specifically, one utilizing the Ad hoc On-Demand Distance Vector (AODV) protocol—in terms of efficiency and timeliness in delivering broker update notifications. Two network mobility scenarios are considered to examine the effectiveness of these strategies under varying conditions: Scenario A (Moderate Mobility) and Scenario B (High Mobility of the Subscriber). In both scenarios, the broker replica migration feature is activated to ensure that broker handoffs take place; the difference lies in the mechanism by which the Publisher and Subscriber are informed of the new Broker’s location. In the baseline method, broker migration updates are disseminated through the AODV protocol, which relies on route discovery and broadcasting mechanisms characteristic of MANET routing. In contrast, the NDN interest-based approach involves the new Broker issuing unicast notifications to the Publisher and utilizing Named Data Networking interest packets directed toward Subscribers. These interest packets follow reverse paths established by prior subscriptions, thereby targeting only the interested Subscribers. Scenario A simulates a moderately mobile Subscriber (moving at approximately 2 m/s over a distance of ∼600 m), reflecting relatively stable topologies, whereas Scenario B introduces a more rapidly mobile Subscriber (3 m/s over ∼900 m), resulting in frequent and dynamic topology changes.

Scenario A—Moderate Mobility (2 m/s): In this scenario, the Subscriber’s movement is limited such that it stays within roughly 600 m of its start point. We observed that under these conditions, the NDN-based notification strategy operates efficiently without requiring any network-wide broadcasts. In particular, the Subscriber’s periodic interest messages (arising from its content requests) provide sufficient state information in the network to track the Subscriber’s location as the Broker migrates. As a result, no global flooding is triggered in the NDN approach, and the new Broker’s location is quickly and efficiently communicated to both the Publisher and the Subscriber. Comparative evaluations between the NDN strategy and AODV in this scenario show that the NDN interest-based mechanism achieves consistently lower broker notification delay. That is, the time to inform the Publisher of the broker change and to inform the Subscriber of the new broker is shorter with the NDN approach than with the AODV-based method. This confirms the expectation that NDN’s stateful forwarding can disseminate migration updates more rapidly. Figure 7 illustrates these results, where the NDN method outperforms AODV in terms of notification delay for the moderate mobility case.

Scenario B—High Mobility (3 m/s): In the higher-mobility scenario, the Subscriber travels a total of about 900 m, causing frequent topology changes. Under these more challenging conditions, we found that the NDN-based strategy may eventually resort to its fallback broadcast notification mechanism. Specifically, once the Subscriber moves beyond the range where previous interest state is effective, the new Broker will trigger a network-wide broadcast to reach any remaining subscribers. Despite this additional overhead, the NDN approach still significantly outperforms the AODV-based approach in terms of notification delay. Figure 8 shows that even though a flooding occurs in the NDN strategy at 3 m/s, the notification delays with NDN remain consistently lower than those with AODV. In our experiments, the AODV-based method struggled to maintain timely updates as the network topology rapidly changed, whereas the NDN method’s adaptive combination of targeted interest notifications (and only occasional broadcast) maintained better responsiveness. These findings indicate that NDN’s intrinsic stateful forwarding and adaptive notification mechanism enhance network resilience and responsiveness under high mobility, yielding superior notification efficiency compared to the traditional AODV protocol. In summary, across both mobility scenarios, the interest-driven notification strategy delivered broker migration updates faster and more reliably, thereby minimizing disruption during broker handoffs.

5.3. Performance Evaluation of DTN7 in High-Mobility Subscriber Scenarios

In earlier experiments, the Ad hoc On-Demand Distance Vector (AODV) protocol demonstrated significant performance degradation, particularly evident at subscriber speeds of 3 m/s and above. To address these limitations, the Delay-Tolerant Networking version 7 (DTN7) protocol was evaluated, yielding substantial enhancements in both message delivery rate and throughput. This evaluation compares two scenarios: Scenario A, which employs baseline AODV-only routing without a DTN layer, and Scenario B, which incorporates DTN7-based routing for content delivery. In both scenarios, broker replication/migration and NDN-based notification mechanisms are assumed to be active as previously described; the distinguishing factor in Scenario B is the deployment of the DTN7 bundle protocol layer on each node. This layer enables nodes to store and carry messages during periods of end-to-end connectivity disruption. The analysis focuses on extreme mobility conditions—for instance, subscriber movement at speeds of 3 m/s or greater—which, as demonstrated in prior experiments, lead to significant communication breakdowns when relying solely on AODV. The evaluation considers message delivery success rate and throughput as key performance indicators to assess the extent to which DTN7 enhances reliability and network capacity under such challenging conditions.

The results clearly show that the integration of DTN7 dramatically improves multicast performance under high mobility.

Table 2. Delivery rates of DTN7 and AODV at various subscriber speeds.

Subscriber Speed (m/s)	DTN7 Delivery Rate (%)	AODV Delivery Rate (%)
3.0	77.96	52.33
4.0	67.50	30.70
5.0	57.70	15.30

depicts the message delivery rates achieved by AODV and DTN7 at various subscriber speeds. DTN7 consistently attains a much higher delivery ratio than AODV at all evaluated speeds. For example, when the Subscriber moves at 3 m/s, the delivery rate with DTN7 is about 77.96 percent, significantly higher than AODV’s 52.33 percent. This advantage becomes even more pronounced as mobility increases: at 4 m/s the DTN7 approach delivers roughly 67.5 percent of messages vs. only 30.7 percent with AODV, and at 5 m/s DTN7 still achieves 57.7 percent delivery while AODV plummets to around 15.3 percent.

Table 3. Throughput of DTN7 and AODV at various subscriber speeds.

Subscriber Speed (m/s)	DTN7 Throughput (msgs/s)	AODV Throughput (msgs/s)
3.0	38.98	26.16
4.0	33.76	15.34
5.0	28.86	7.67

compares the average throughput (in messages per second) between DTN7 and AODV in the same scenarios, which mirrors the delivery rate trends. At 3 m/s, DTN7 sustains an average throughput of about 38.98 msgs/s, notably higher than the 26.16 msgs/s with AODV. At 4 m/s, DTN7 delivers 33.76 msgs/s vs. 15.34 msgs/s for AODV, and at 5 m/s, DTN7 still manages 28.86 msgs/s whereas AODV collapses to roughly 7.67 msgs/s. These results indicate that, without DTN7, the network’s capacity to deliver content drops sharply at high mobility, but with DTN7 the throughput remains much more stable.

The improvements provided by DTN7 can be attributed to its inherent design tailored for intermittent connectivity and frequent network fragmentation. DTN7 utilizes a store–carry–forward mechanism, which allows nodes to temporarily store messages when immediate forwarding paths are unavailable, subsequently transmitting these messages when connectivity is restored. This approach significantly reduces packet losses that typically occur due to frequent topology changes and unstable routes, which are common in highly dynamic mobile environments. Consequently, DTN7 effectively increases both the reliability of message delivery and the overall throughput.

These experimental outcomes confirm that integrating DTN7 significantly mitigates the performance degradation experienced by AODV at higher mobility levels. By effectively managing intermittent connectivity and network fragmentation, DTN7 maintains considerably higher message delivery rates and throughput under challenging mobility scenarios, underscoring its efficacy as a robust solution for highly dynamic mobile ad hoc networks.

6. Related Work

6.1. Support of Node Mobility in Pub/Sub Systems

Pub/Sub enables multicasting that is resilient to changes in the logical structure of the network. Content-Based Multicast (CBM) [9], described in Section 2, also addresses the challenge of network topology. The primary distinction between Pub/Sub and CBM is the explicit use of brokers, which CBM does not employ despite utilizing similar mechanisms. For Pub/Sub systems to function effectively in MANET, where participating nodes can move physically, a mechanism that facilitates connection handover between nodes in response to such movements is crucial [22].

Huang and Garcia-Molina [23] concentrated on supporting node mobility in Pub/Sub systems and discussed strategies for distributing the load of Pub/Sub processing across multiple nodes, as well as replicating the system to enhance resilience to faults and message loss due to node mobility. In [24], the same research group explored optimization techniques for the Pub/Sub tree (PST), which is responsible for message delivery from the source to the receiver. They introduced a novel metric to evaluate the quality of PSTs, considering both Pub/Sub objectives and MANET characteristics, and proposed a distributed algorithm called SHOP PARENT that constructs PSTs greedily. Costa et al. [25] proposed a Pub/Sub routing framework named SocialCast, which predicts node mobility based on social interaction metrics and performs efficient routing using these mobility predictions. The framework’s effectiveness was evaluated using mobility models derived from real mobility traces. Amozarrain and Larrea [26] proposed MFT-PubSub, a content-based Pub/Sub protocol for MANET. This reactive method supports node mobility in Pub/Sub systems and can handle the physical and logical mobility of both clients and brokers in MANET without prior knowledge of the nodes’ connections. Experimental results show that MFT-PubSub achieves a higher message delivery rate than AODV.

The impact of node mobility on Pub/Sub systems has also been analyzed experimentally. Burcea et al. [27] formalized the support for node mobility in Pub/Sub systems and developed methods to minimize the associated costs. As a result, they discovered that unicast, rather than multicast, is necessary to support mobile users, leading to a doubling of the required network capacity if 10% of the users are mobile. Additionally, they developed and evaluated techniques to reduce the cost of supporting source mobility [28], demonstrating that the use of appropriate routing protocols can mitigate the degradation of scalability.

6.2. Mobile Pub/Sub Using Information-Centric Networks

Another promising approach for supporting the physical mobility of nodes in Pub/Sub systems is leveraging Information-Centric Networks (ICNs). Several studies have explored this direction: Xylomenos et al. [29] proposed an ICN architecture called Publish–Subscribe Internet (PSI). They provided an overview of PSI, covering network bootstrapping to information delivery, and emphasized how PSI addresses Internet shortcomings, particularly in network-layer caching and mobility. Detti et al. [30] examined the advantages of implementing Pub/Sub systems on MANETs using Content-Centric Networking (CCN). Their findings highlighted that CCN features such as in-network caching and multicasting can facilitate efficient and reliable data delivery in MANET environments. The performance evaluation was conducted using the CCN MANET routing engine, developed as a plug-in for the OLSR Linux daemon in an emulation environment. Ventrella et al. [31] proposed pull-based and push-based Pub/Sub communication schemes designed to support user mobility over ICN. They developed an analytical model to describe the communication overhead associated with these schemes and validated its accuracy through simulation.

7. Concluding Remarks

This paper focuses on Pub/Sub multicast over MANET, proposing a method to enhance resilience to node mobility and evaluating it using the ns-3 simulator. The core concepts of the proposed method are as follows: (1) utilizing DTN7 at the network layer (bundle layer); (2) ensuring the continuity of the broker function by dynamically generating and synchronizing broker replicas; and (3) maintaining communication paths by having the broker and its replicas periodically broadcast advertisement packets—a technique inspired by Named Data Networking (NDN). Simulation results show that the proposed approach significantly reduces message delivery latency and improves overall throughput.

Future work includes conducting more extensive simulations, including scalability evaluations, and deploying the proposed method by constructing a smartphone-based MANET to assess its effectiveness in a real-world environment.