An overview of the existing literature with research focused on the SIP signaling performance over MANET is presented in this chapter. In addition, an extensive survey is provided on the related work. This review is focused on the studies about SIP signaling over MANET and the performance enhancement approaches for SIP-based VoIP applications. Generally, SIP is implemented over MANETs with four different types of SIP signaling systems as represented in Figure 3-1. The first of them is peer-to-peer SIP over MANET. The main purpose in this case is not to use SIP servers. This chapter explains in detail much more about this kind of system in the following review of the existing research in this direction. The second type of SIP signaling system is SIP with multiple servers over MANET that includes registration, redirect and proxy. The third type of SIP signaling system that exists and is researched is SIP with a single SIP server that acts as a registration, redirect and proxy server over MANET. The research work in the presented thesis is focused on this kind of signaling system. The fourth type of SIP signaling system over MANET is SIP with an IMS system. This chapter will provide in the following exposition the current state of the research, results, gaps, advantages and disadvantages regarding the above-mentioned SIP signaling systems over MANET. Furthermore, the available performance enhancement methods for SIP signaling over MANET will be discussed in the current chapter.
Figure 3-1: Literature survey of types of SIP signaling system implementations over MANET
Peer-to-Peer [85, 86, 87, 89,
90, 96]
SIP over IMS Servers [104, 105, 106, 107] SIP Signaling Systems over MANET
Single SIP Server [100, 101, 102,
103] Multiple SIP Servers
49 The four implementation types for SIP signaling systems had been introduced in Chapter 2. In this chapter, a survey about the SIP signaling system implementations over MANET will be provided to have a comprehensive review of the current state of the art. There are a number of researchers who focused on adapting the SIP to MANETs. Generally, they can be classified in two classes, according to which of the nodes take action as SIP servers in the network. The first class is characterized with the implementation of the SIP servers to all nodes and each node is registering locally or is broadcasting the location information in the entire network. The other class is distinguishing some nodes to take action as SIP servers. This literature review represents the current state of the proposed research area in terms of the investigation, the evaluation and the service enhancement efforts. Moreover, the supportive simulation tools and test-beds for SIP signaling implementations for MANET will be discussed in this chapter.
3.1 Peer-to-Peer SIP Signaling Implementations over MANET
The authors of [83] propose two solutions for enabling SIP in MANETs: dSIP and sSIP. According to the dSIP, a REGISTER request is broadcasted from each node in order to notify all nodes in the network with the information of its location. Discovery of the members in the network is accomplished by local probing in the cache. To enable Session Initiation Protocol in MANETs, the Service Location Protocol (SLP) [84] is used by sSIP [83]. An SLP request is broadcasted from the node that is connecting with the ad hoc networks in order to ask for bindings of the users that are available. After this, every node that receives an SLP request responds using an SLP reply that includes its binding. However, as mentioned earlier, using this kind of solution can cause flooding. That is the reason for problems arising when they are used in larger ad hoc networks.
The authors of [85] employ peer-to-peer cover that is structured and related to Chord [86]. With the purpose of mapping the users with the relevant connection information, a Distributed Hash Table (DHT) protocol is used by the nodes. Hence, when some of the nodes connect to the Chord cover, they will be in charge for keeping the information related with the part of the cover that is mapping to its estimated Node-Identification (Node-ID). High control overhead is caused by maintaining the hash tables. Registration in [87] is realized utilizing the multicast mechanism with IPv6. A REGISTER request is multicasted by the node to announce its appearance in the whole network. User-List-Cache is updated by each node when REGISTER updates are received. After that, each node is replying by providing the information to the correspondent employing unicast. However, this kind of alternative is also not effective because, for large ad hoc networks, preserving the User-List-Cache gives poor results. Initial research on the subject of SIP over MANET was started in 2003 by Khlifi et al [88]. In this work, a framework for conference signaling using SIP is presented allowing a MANET user to discover, initiate conferences, and join existing conferences with other users. After this work, the next research
50 on SIP over MANET was in 2004 in [89], where SIP is set up over OLSR utilizing a cross layer, integrated application and routing layer, in order to assist proxy-less and proxy-based systems. A proxy- less system is without a proxy server and proxy-based SIP MANET contains at least one SIP proxy server. Research in the field of proxy-less SIP MANET, i.e. SIP peer-to-peer over MANET without SIP servers is presented in [90]. The authors in this work proposed a signaling system that is unique and is used for sessions in P2P ad hoc networks. The framework when SIP is used in ad-hoc networks provided in [88] is enhanced in [90] by establishing hierarchical clustering architecture. This concept is confirmed by performing testbed running on eight computers. The benefit of the system proposed in [90] is in generating a lower number of overhead messages compared to [88].
The majority of the P2P SIP over MANET approaches employs resource discovery mechanisms in order to offer SIP user location discovery. Hence, P2P SIP over MANET approaches could be also classified into P2P SIP without overlay network and on the overlay network. Although the majority of the SIP over MANET approaches employs SIP register and user discovery operations in MANET, they do not deal with the working of their protocols in heterogeneous networks in order to supply interoperability between MANET and Internet SIP users. Research with solutions for Internet connected MANET environments are presented in [83, 91, 92, and 93]. Yet, this research relies on a centralized SIP registrar/proxy that can be positioned at the Internet or at the MANET gateway. However, this kind of centralized registrar/proxy is the reason for creating a traffic bottleneck when SIP requests are sent to the gateway and for creating a single point of failure. With centralized architecture, users’ SIP binding information is kept on MANET by one or a few MANET gateways, named SIP gateways. The role of the SIP gateways is also to forward the received SIP register requests from MANET users to an external SIP registrar on Internet.
In [87] the authors design and implement the pseudo Session Initiation Protocol (p-SIP) server embedded in each mobile node in order to provide the ad-hoc VoIP services. The benefit from this paper is that the implemented p-SIP server is compatible with common VoIP user agents, and it integrates the standard SIP protocol with SIP presence in order to handle SIP signaling and discovery mechanism in the ad-hoc VoIP networks. The advance of this work is also that the implementation is based on real equipment. The implementation of p-SIP is done on IBM ThickPAD x32 laptops, equipped with IEEE 802.11g wireless communication, using the Ubuntu Linux 6.10 and applying Kphone 4.2 as UA on top of the embedded p-SIP server. These authors, with the implementation of the testbed and the performance measurements from the experimental setup, have shown valuable analysis of the ad-hoc VoIP network. The results of this work demonstrate the possibility to achieve ad-hoc VoIP services using the implemented p-SIP servers. However, in this work more complete presentations of different UDP packet sizes, injection rates and contention scenarios are not provided. What is provided is the information on the influence of TCP/UDP traffic that contend VoIP streams in ad-hoc networks. Further possible research in the direction of the work in [87] is the influence of the ad-hoc nodes’ density on
51 the performance and the limitation of forwarding hop counts to realize acceptable VoIP QoS in the ad- hoc network.
The authors of [94] suggest a framework for service provisioning in stand-alone MANETs. Contributions that this work provides are in the new model of business that is harmonized with the features of MANETs and that allows invocation and execution of the services, and in the allocation system of the SIP servlets and overlay networks as a service execution environment. Because the proposed model does not have a central unit, and it is adaptable when dedicating functions in order to grant the functionalities, any user can take part in possessing the required features. The suggested functional distribution by the authors of this work deals with the number of independent units and the loose coupling. In [94] the authors also propose a covering network for execution of the services in MANETs that are stand-alone, based on the framework of the SIP servlets. The benefit of the paper is also in the prototypes built as verification of ideas for the model of business and the allocated system. Provided solutions in this work are proving that the model and the scheme are reasonable with a satisfactory time of response. The covering network protocol in the results is also formally validated. However, yet more detailed validation would be needed here in this presented work.
In [95] these authors propose the architecture of a MANET emulator suitable for SIP services on a MANET. The proposed architecture in this work supports real-time audio/video communication, node mobility, and peer-to-peer-type communication. The SIP_MANET emulator has been developed in [95] based on the proposed architecture, and it is confirmed that solid communication quality can be maintained with SIP applications. Communication quality evaluation is also conducted to confirm the effectiveness of the simulator. In order to make achievable usage of the MANET emulator for verifying a SIP application, it is suggested to add the capabilities to translate the IP address and port numbers, to give priority to AODV packets, and to process transmission/reception of packets in multiple threads. The percentage of successful audio and video communication in a SIP application is approximately 95% when the nodes stay still, hence the communication quality in this case is satisfactory. From the given results in this work when the nodes are moving, this percentage is decreased to approximately 77%. Still, in the presented work in [95] multi-path protocols are not taken into consideration and are not included in the testing simulations. Hence, to enhance the communication quality when nodes are in movement, additional research of multi-path protocols is needed here.
The work in [96] represents an innovative Peer-to-Peer (P2P) framework for SIP on MANET. The focus here is on distributed P2P resource lookup mechanisms for SIP that tolerate failures resulting from the node mobility. The authors of [96] proposed a novel P2P lookup architecture based on a Structured Mesh Overlay Network (SMON) that enables P2P applications to perform fast resource lookups in the MANET environment. Their approach extends the traditional SIP user location discovery utilizing DHT in SMON in order to distribute SIP object identifiers over SMON. Simulation and experiment results
52 are conducted from which it is concluded that SIPMON provides the lowest call setup delay comparing it with the existing broadcast-based approaches. In [96] the authors also propose a new OLSR Overlay Network (OON), which is a single overlay network that contains MANET nodes and nodes on the Internet. Testbed experiments are conducted proving that extended SIPMON (SIPMON+) gives better performance expressed in terms of call setup delay and handoff delay compared with MANET for Network Mobility. These authors also contribute in [96] with a proof-of-concept and prototype of P2P multimedia communication based on SIPMON+ for post-disaster recovery missions. This concept is evaluated with experimentation in real disaster situations – Vehicle to Infrastructure scenarios - and it is concluded that the proposed prototype outperforms MANEMO-based approaches expressed in packet loss, call setup delay, and deployment time. The proposed framework in [96] can be easily implemented to day-to-day growth of the Internet connectivity. It will be interesting to see this work to be continued in addressing how TCP-based applications can be provided on SIPMON+. In this direction, other mobility issues like session mobility are motivating for research.
3.2 Implementations of Multiple SIP Servers over MANET
The authors of [97] propose AdSIP, that is a protocol that is distributed and that allows SIP in MANETs. Realization of this protocol is on the network simulator ns-2 where comparison is made with the Tightly Coupled Approach (TCA) in terms of pertinent metrics, average session establishment time, failure rate and consumed bandwidth. Comparison shows that the proposed protocol in [97] performs improved adaptability and scalability to the mobility of the nodes. The solution in this work called AdSIP, chooses a group of nodes that are mobile to operate as SIP servers, and they establish a virtual infrastructure as overlay on top of the physical network. A new distributed algorithm is built to construct the topology and to assign dynamically previously explained functionality to a group of nodes in the network. The simulation results obtained using the ns-2 simulator clearly state that the proposed AdSIP protocol is well-adapted to mobile ad hoc networks, giving lower session establishment time, low control overhead and high service availability. However, this work lacks confirmation of the proposals with real results that could be obtained with a real scenario, instead of the ns-2 simulation tool.
Proactive route optimization in SIP mobility is introduced in [98] in order to achieve a reduction in the session setup latency. According to the proposed Session Initiation Protocol – Proactive Route Optimization (SIP-PRO), during the location registration step the mobility binding information is pre- fetched and used for session establishment. Reduced session setup latency in this work is accomplished by eliminating traverse over multiple SIP servers using the proactive route optimization. Hence, in this work these authors propose a novel idea called proactive route optimization. It achieves, when a session is initiated, direct establishment of the session with the callee if the caller has valid mobility binding information. A mobility-aware pre-fetching scheme is developed where only the mobility binding
53 information with lower mobility is selected, as it is more probable that such information could be used for session establishment. In [99] the authors also propose a new session setup procedure where mobility information with a sufficient residual time is used. Still, this work fails to perform extensive simulations with the developed analytical models in order to verify the proposed optimization and procedures.
3.3 Implementations of Single SIP Server over MANET
The implementations of Single SIP Server over MANET will be used in this research study as will introduced in Chapter 6. The overview for the current state of the art will help with understanding the SIP implementation issues and available efforts to reference the efforts of this research study with the related researches in the literature. In [100], An intelligent VoIP system with embedded pseudo SIP server in an ad-hoc network is implemented. The embedded pseudo SIP server presented in this work is compatible with common VoIP user agents using SIP and it acts like middleware between the application and the transport layer. The VoIP quality level of the service is performed with the transmission delay for signaling and voice packets and it shows acceptable results with the conducted testbed. The proposed pseudo SIP server in [100] utilizes SIP presence to discover the mobile device and exchange the signaling over an ad-hoc network. However, this work lacks other performance metrics (not just transmission delay) in the experimental results to confirm the quality of the proposed SIP server.
Converting IP addresses and port number and rewriting SIP messages is required in order to enable a MANET emulator to provide SIP services. However, the disruptions may arise between SIP clients, and the real-time performance can fall. The authors of [101] propose an architecture of a MANET emulator and local multipath routing appropriate for SIP services. A SIP_MANET emulator is developed and correct operation of the SIP_VoIP call is confirmed. The proposed routing method from the authors of this work provides high probability of retaining the required path. Their developed system is well described and the evaluation results are presented in detail. The proposed routing method is compared with AODV and the disjoint multipath routing, using the MANET emulator and the described evaluation model. Call holding time is measured, defined as the time from the start to the disconnection of the call. Path retaining probability is also calculated and it verifies the effectiveness of the proposed local multipath routing. The proposed routing method uses a spare path when some node in the used path fails and that is the reason why its path retaining probabilities are higher than that of AODV. It would be very useful if the proposed local multipath routing in [101] is compared with AODV on more varied network models to have more detailed results in this domain.
A SIP-based mobile network architecture for Network Mobility (NEMO) in vehicular applications is developed in [100]. The focus in this work is on developing a MANET that contains hosts that are mobile and that are in a vehicle or in a group of vehicles. In [100] the MANET is linked with a SIP-
54 based Mobile Network Gateway (SIP-MNG) to the outside, equipped with external wireless interfaces and internal 802.11 interfaces. In this direction, SIP-MNG is supporting call admission control and resource management for the MHs. The authors are proposing a boost mechanism with message service that is short for the purpose of waking up in an on-demand manner the wireless interfaces. Signaling with details is presented regarding this mechanism. Additionally, this system is completely well- matched with the SIP standards that are accessible. Prototyping practice and performance measurements outcomes are presented. The proposed system in this work saves fees for internet access that is beneficial for operators of public transport or users, allowing the sharing of one interface from multiple sessions. Furthermore, with this kind of design, vehicles could offer access to the Internet to travellers supporting the group mobility. A proposed push mechanism allows SIP-MNG to stay off-line when calling activity is not present and to be activated when there is a need. Maintaining global accessibility of the users, the proposed push approach also saves call charges and energy. From the presented experimental results, it is demonstrated that for PHS, WCDMA, and 802.11 networks, it is possible for