Comparative Analysis of the Extended Persistence Policy on Distance-Vector and Link-State Routing Performance in the DOUNG Ad Hoc Network

1. Introduction

Mobile ad hoc networks (MANETs) are infrastructureless wireless networks in which each node may act simultaneously as a sender, receiver and router. This property is attractive where fixed infrastructure is unavailable, expensive or insufficiently dense. It is also relevant for distributed educational systems in which learners, teachers, mobile laboratories and local servers are spread over a campus or over weakly connected sites. In the New Generation of Open Digital Universities (DOUNG), an ad hoc network can support document access, content synchronization, messaging and collaborative learning without permanent dependence on a central access point.

The main challenge is topology instability. A route that is valid at one instant may disappear shortly after because a node moves, a link weakens, a relay leaves radio range or the medium becomes temporarily congested. In a MANET, packet loss may therefore result from congestion, route repair, radio interference, channel contention or delayed route-table convergence. This ambiguity is critical for reliable transport because the transport layer reacts to missing acknowledgments without directly observing the network-layer state. SCTP provides reliable transport over IP, but its acknowledgment and retransmission mechanisms may become inefficient when the transport layer does not know the real routing state [1].

MANET routing protocols can be grouped into several families. AODV is a reactive distance-vector routing protocol designed for mobile ad hoc networks, using on-demand route discovery and sequence numbers to avoid loops [2]. DSR is based on source routing and route discovery and maintenance mechanisms [3]. OLSR represents the link-state family and reduces flooding overhead through multipoint relays [4]. DSDV is a proactive distance-vector protocol based on routing-table updates and sequence numbers [5]. Recent MANET routing-protocol evaluations also emphasize that mobility, traffic load and heterogeneous network conditions strongly affect comparative performance [9,10,11].

The extended persistence timeout policy, denoted VTP2, follows a cross-layer design. It delays SCTP retransmissions when the network layer indicates that the route is unavailable. The principle is that the transport layer should not systematically interpret the absence of acknowledgment as congestion. Previous work on SCTP/DSDV highlighted the interest of such a policy for avoiding unnecessary retransmissions when the route or channel is temporarily degraded [6].

The contribution of this paper is threefold. First, it formalizes the application of VTP2 in the DOUNG ad hoc network. Second, it compares the effect of VTP2 on distance-vector/reactive protocols (AODV, DSDV and DSR) and on the link-state protocol OLSR. Third, it presents the results by protocol, followed by a comparative discussion and a final global synthesis.

2. Materials and Methods

2.1. Network Model

The DOUNG ad hoc network is modeled as a dynamic wireless graph G(t) = (V(t), E(t)), where V(t) is the set of active nodes and E(t) is the set of available radio links at time t. A node may represent a learner terminal, a teacher device, a mobile access unit, a local educational server or a relay. A link belongs to E(t) when two nodes can exchange frames under the current radio and mobility conditions. Because the network is mobile, both V(t) and E(t) vary during the simulation period.

A route Rsd(t) between a source s and a destination d is considered valid when all intermediate links required by the selected path are available at time t. If one of these links disappears, the route becomes unavailable until the routing protocol discovers, repairs or computes a new usable path. This route-unavailability interval is the key situation addressed by VTP2. During such an interval, a transport-layer retransmission is usually inefficient because the problem is not the loss of one isolated packet but the absence of a complete forwarding path.

To capture this behavior, we define a binary route-state variable delta_R(t). The variable is equal to 1 when a route between the communicating endpoints is available and equal to 0 when the route is unavailable. This simplified representation is useful because it separates two phenomena that are often mixed at the transport layer: packet loss due to congestion and packet loss due to route interruption. VTP2 uses this state as a condition before authorizing retransmission.

2.2. Classical SCTP Retransmission and VTP2 Principle

In classical SCTP, retransmission is driven by the expiration of the retransmission timeout TRTO and by acknowledgment behavior [1]. This is appropriate when the path is stable and packet loss mainly reflects congestion or random errors. In a MANET, however, a timeout may expire because the route is temporarily broken. If SCTP retransmits immediately, the retransmission may be queued, dropped or propagated over an incomplete path. Repeated retransmissions can also increase medium access contention, which is harmful for neighboring flows.

VTP2 modifies this behavior by adding a route-availability condition to the retransmission decision. Retransmission is allowed only if TRTO expires and the route-state variable indicates that the path is usable. The decision can be represented as follows:

RTX_VTP2(t) = 1 if TRTO expires and delta_R(t) = 1; RTX_VTP2(t) = 0 if delta_R(t) = 0.

(1)

This expression does not change the reliability objective of SCTP. It changes the timing of retransmission in order to avoid sending when the routing layer is unable to forward the packet. In practice, VTP2 acts as an extended persistence mechanism: the sender persists in waiting while the route is unavailable, then resumes retransmission when the network layer indicates that communication can continue. The policy is therefore especially relevant for routing protocols that experience explicit route discovery, route maintenance or convergence phases.

The VTP2 mechanism belongs to the broader family of cross-layer approaches for ad hoc networks [7]. Cross-layer design must be used carefully because excessive coupling between protocol layers can reduce modularity. In this work, the coupling remains limited and interpretable. The transport layer receives only route-availability information, not the full routing table or internal routing logic. This keeps the mechanism simple while still addressing the main ambiguity that affects SCTP behavior in mobile ad hoc networks.

2.3. Routing Protocols Considered

Four routing protocols are considered in order to compare different routing families. AODV is a reactive distance-vector routing protocol. It discovers routes on demand and uses sequence numbers to avoid routing loops [2]. When a route is needed, the source initiates route discovery. This behavior limits control overhead when traffic is sparse but may introduce discovery delays when mobility breaks an active path.

DSR is also reactive, but it relies on source routing and route caches [3]. The complete route is carried in packet headers, and the protocol includes route discovery and route maintenance mechanisms. DSR can exploit cached paths, but stale routes may appear under mobility. This makes it sensitive to the timing of transport retransmissions when the cached route is no longer valid.

DSDV is a proactive distance-vector routing protocol based on periodic routing-table updates and sequence numbers [5]. Unlike AODV and DSR, it attempts to maintain routes before data packets need them. Nevertheless, table convergence is not instantaneous. During topology changes, a node may temporarily hold outdated or incomplete information. This creates transient periods in which VTP2 can still prevent ineffective retransmissions.

OLSR is a proactive link-state routing protocol [4]. It reduces flooding overhead through multipoint relays and maintains topological information that can be used to compute routes. Because OLSR proactively disseminates topology information, it is expected to experience fewer complete route-unavailability periods than purely reactive protocols. However, it still requires time to reconverge after mobility events, and it consumes energy and bandwidth through control messages. Comparing OLSR with AODV, DSDV and DSR helps determine whether VTP2 is mainly useful for route-discovery protocols or also beneficial for proactive link-state routing.

2.4. Simulation Parameters

The experiments are organized in an NS-3-compatible simulation environment using IEEE 802.11 at the MAC layer, SCTP at the transport layer and four routing protocols: AODV, DSDV, DSR and OLSR [8]. Node mobility is varied from 0 to 20 m/s to reproduce static classrooms, mobile campus areas and highly dynamic digital-university scenarios.

Table 1. Simulation parameters.

Parameter	Value
Network type	MANET/DOUNG ad hoc network
Simulation area	1000 × 1000 m2
Number of nodes	50 nodes
Mobility model	Random Waypoint
Node mobility	0, 5, 10, 15 and 20 m/s
MAC layer	IEEE 802.11
Transport	Classical SCTP and SCTP with VTP2
Routing protocols	AODV, DSDV, DSR, OLSR
Simulation duration	300 s
Metrics	PDR, throughput, delay, SCTP retransmissions, energy

Table 1. Simulation parameters.

Parameter	Value
Network type	MANET/DOUNG ad hoc network
Simulation area	1000 × 1000 m2
Number of nodes	50 nodes
Mobility model	Random Waypoint
Node mobility	0, 5, 10, 15 and 20 m/s
MAC layer	IEEE 802.11
Transport	Classical SCTP and SCTP with VTP2
Routing protocols	AODV, DSDV, DSR, OLSR
Simulation duration	300 s
Metrics	PDR, throughput, delay, SCTP retransmissions, energy

2.5. Performance Metrics

Five metrics are used to evaluate the effect of VTP2. The packet delivery ratio (PDR) is defined as PDR = (Nreceived/Nsent) × 100. Throughput is computed from the amount of data successfully received during the useful simulation period. The average end-to-end delay is the mean difference between packet reception and transmission times. The SCTP retransmission rate measures the proportion of retransmitted packets. Finally, average energy consumption is estimated from radio activity and avoided retransmissions.

A global VTP2 gain index is used to synthesize the results. It combines the average gains in PDR and throughput with the average reductions in delay, retransmissions and energy consumption. This index is not intended to replace individual metrics; it only provides a common scale for protocol comparison.

3. Results

3.1. General Behavior Under Mobility

Across all protocols, mobility reduces performance because it increases the probability of route breaks, neighborhood changes and transient routing inconsistencies. Under classical SCTP, the transport layer continues to react mainly through timeout-based retransmission. This behavior becomes less efficient as mobility increases because the transport layer may retransmit during periods when the network layer has not yet restored a usable route.

With VTP2, the degradation remains present but is less pronounced. The policy does not eliminate route breaks and does not replace routing. Its benefit comes from better synchronization between the transport and network layers. When a route is unavailable, VTP2 prevents immediate retransmission and therefore avoids part of the radio activity that would not contribute to successful delivery. The magnitude of the improvement depends on how often the routing protocol enters route discovery, repair or reconvergence phases.

3.2. AODV Results

AODV benefits strongly from VTP2 when mobility increases. Since AODV discovers routes on demand, a route break is followed by a discovery or repair interval during which data forwarding may be temporarily impossible. Classical SCTP may interpret missing acknowledgments during this interval as a sign that retransmission is required, even though the main limitation is the absence of a valid path. VTP2 reduces this mismatch by postponing retransmissions until the route becomes available again.

The reference results show that the average throughput gain for AODV reaches 17.64%, while the average reduction in SCTP retransmissions reaches 30.30%. The PDR gain is 8.41%, and the delay reduction reaches 15.38%. These values show that VTP2 not only reduces retransmissions but also improves useful delivery. The energy reduction of 13.88% is consistent with lower radio activity. For DOUNG, this is important because AODV-like behavior is attractive where continuous maintenance of complete routing information would be costly.

Figure 1. AODV results: classical SCTP versus VTP2 under increasing mobility.

Figure 1. AODV results: classical SCTP versus VTP2 under increasing mobility.

3.3. DSDV Results

DSDV obtains a clear but slightly lower benefit than AODV. As a proactive distance-vector protocol, DSDV maintains routing tables before data transmission. This can reduce route-discovery delay, but it does not fully suppress transient inconsistencies. Under mobility, routing-table entries may temporarily describe paths that are no longer valid or may require several updates before reflecting the current topology. During these intervals, classical SCTP may still retransmit prematurely.

With VTP2, unnecessary retransmissions are reduced while routing information converges. The average PDR gain is 6.34%, the throughput gain is 13.52% and the retransmission reduction is 28.94%. The delay reduction of 13.49% shows that waiting for route availability can be more efficient than repeated transmission over unstable paths. The energy reduction of 10.61% remains substantial. DSDV therefore occupies an intermediate position: proactive operation helps, but distance-vector convergence still creates cases where extended persistence is useful.

Figure 2. DSDV results: classical SCTP versus VTP2 under increasing mobility.

Figure 2. DSDV results: classical SCTP versus VTP2 under increasing mobility.

3.4. DSR Results

DSR exhibits behavior close to AODV because it is also reactive and strongly affected by route discovery and route maintenance. Its source-routing mechanism can be efficient when cached routes remain valid. However, under mobility, cached routes may become stale, and packets may be sent using paths that no longer exist. When acknowledgments are delayed or missing because of such path invalidity, classical SCTP may retransmit before the routing layer has completed repair.

VTP2 improves DSR by preventing SCTP from retransmitting while the source route is unavailable. The average PDR gain is 8.30%, close to AODV. The throughput gain is 17.53%, and the retransmission reduction is 28.30%. The delay reduction reaches 15.03%, confirming that postponed retransmission can shorten effective delivery when repeated failed attempts are avoided. The energy reduction of 13.18% again links route-aware retransmission with lower radio activity. DSR with VTP2 is therefore useful when route caching is beneficial but must be protected against stale-route effects.

Figure 3. DSR results: classical SCTP versus VTP2 under increasing mobility.

Figure 3. DSR results: classical SCTP versus VTP2 under increasing mobility.

3.5. OLSR Results

OLSR benefits from VTP2, but the improvement is more moderate. This result is coherent with its proactive link-state design. OLSR maintains topology information and computes routes without waiting for a data packet to trigger discovery. Therefore, complete route-unavailability periods are generally less frequent than in reactive protocols. Moreover, the multipoint relay mechanism reduces the cost of flooding link-state information [4].

Nevertheless, OLSR is not immune to mobility. Topology information must be disseminated, received and processed before all nodes converge. During reconvergence, VTP2 can prevent some ineffective SCTP retransmissions. The average PDR gain is 3.41%, the throughput gain is 6.20% and the retransmission reduction is 21.76%. The delay reduction is 8.95%, and the energy reduction is 4.93%. The lower energy gain is expected because OLSR continues to send control messages independently of data retransmissions. VTP2 therefore stabilizes OLSR but has a stronger corrective effect on AODV, DSDV and DSR.

Figure 4. OLSR results: classical SCTP versus VTP2 under increasing mobility.

Figure 4. OLSR results: classical SCTP versus VTP2 under increasing mobility.

Table 2. Average gains obtained with VTP2.

Protocol	PDR gain (%)	Throughput gain (%)	Delay reduction (%)	RTX reduction (%)	Energy reduction (%)
AODV	8.41	17.64	15.38	30.30	13.88
DSDV	6.34	13.52	13.49	28.94	10.61
DSR	8.30	17.53	15.03	28.30	13.18
OLSR	3.41	6.20	8.95	21.76	4.93

Table 2. Average gains obtained with VTP2.

Protocol	PDR gain (%)	Throughput gain (%)	Delay reduction (%)	RTX reduction (%)	Energy reduction (%)
AODV	8.41	17.64	15.38	30.30	13.88
DSDV	6.34	13.52	13.49	28.94	10.61
DSR	8.30	17.53	15.03	28.30	13.18
OLSR	3.41	6.20	8.95	21.76	4.93

4. Discussion

4.1. Distance-Vector/Reactive Routing Versus Link-State Routing

The comparison confirms that VTP2 has a stronger effect on distance-vector/reactive protocols than on OLSR. AODV and DSR are directly exposed to route discovery or repair periods. When the route is unavailable, classical SCTP still reacts to missing acknowledgments by retransmitting, which increases radio contention and can aggravate local congestion. VTP2 introduces synchronization between the transport layer and the network layer: retransmission is postponed as long as the route is not usable.

DSDV occupies an intermediate position. Although it is proactive, it is based on distance-vector tables and may exhibit transient inconsistencies during topology changes. VTP2 then reduces retransmissions while the routing table has not yet converged. This explains why DSDV gains are significant but slightly lower than those obtained with AODV and DSR.

In OLSR, routes are proactively maintained through topology information dissemination and the multipoint relay mechanism. The effect of VTP2 is therefore more moderate: the protocol often has a route before data transmission starts, but it can still suffer during reconvergence. Consequently, VTP2 plays mainly a stabilizing role in OLSR, whereas it plays a corrective role in AODV, DSDV and DSR.

The global index in Figure 5 visually confirms this hierarchy. AODV and DSR obtain the highest indices, followed by DSDV. OLSR has the lowest global gain, not because VTP2 is ineffective, but because its proactive operation already limits some route-unavailability periods.

Figure 5. Global VTP2 gain index by routing protocol.

Figure 5. Global VTP2 gain index by routing protocol.

4.2. Implications for the DOUNG Network

In the DOUNG network, educational flows may include document access, messaging, content synchronization and virtual classrooms. Applications that tolerate moderate delay but require reliable delivery particularly benefit from VTP2. In areas with high mobility or intermittent connectivity, AODV or DSR combined with VTP2 may provide a relevant compromise. In more stable and dense environments, OLSR remains attractive because of its proactive route availability, but its control overhead must be considered.

5. Conclusions

This paper transformed the theoretical analysis of the extended persistence timeout policy into a protocol-oriented experimental study. The results show that VTP2 improves SCTP performance in the DOUNG ad hoc network, especially when the network layer experiences temporary route unavailability.

The highest gains are observed for AODV and DSR, followed by DSDV. OLSR also benefits from VTP2, but to a lesser extent, because its link-state operation proactively maintains routes. The main conclusion is that VTP2 improves protocols that are sensitive to route breaks and route repair phases more than it improves link-state protocols.