<p>We depend more and more on networked systems in our every day lives. The functionality and correctness of such systems are thus becoming increasingly important both for individuals and the society as a whole. At the same time, people expect new services to be developed more rapidly than ever. These two forces makes the design of these often complex services a difficult task. Model-driven techniques are in widespread use for the design of distributed services, and software tools can support the designer through the development process. Our goal is to help the user of such modelling tools making it easier to develop correct services more quickly. When we take a closer look at existing services, we find great similarities in the way two parts of the service interact. For instance, one part might request for some information from another part or perhaps notify another part of something. We identify them as so-called interaction patterns, which describe these generic interactions at a high level of abstraction. They are modelled using UML~2.0 collaboration templates. We have defined an approach for using a number of interaction patterns to describe the interaction interface between two parts, modelled by a UML~2.0 collaboration. The user of the tool selects, customises and applies a number of patterns to the interaction interface. The applied patterns can then be composed, meaning that the user determines their sequence of activation. The composition technique defined is based on the use of UML~2.0 state machines and submachine states. We have designed an algorithm which generates the behaviour of the interaction interface based on the applied patterns and their composition. The behaviour is expressed using two state machines which describes the allowable sequence of signals for each participant in this interaction. To ensure the correctness of the behaviour of the interaction interface, we have identified a number of common design flaws that might occur during the composition of pattern instances based on the work by Jacqueline Floch. The error situations are either prevented through reasonable constraints on the used model elements or resolved when the behaviour of the interaction interface is generated by the defined algorithm. The approach has been implemented as an Eclipse plug-in and integrated with Ramses, a modelling tool suite developed at the Department of Telematics. This provides the possibility to get hands-on experience with using our approach for the design of reactive services. We believe that our work can make it easier for the user of such tools to design correct services more rapidly.</p>

<p>This thesis studies the use of collaborations as a mean to visualize behaviour of an observed distributed reactiv system. The work is inspired by previous work where an approach using elementary collaborations, that is collaborations with only two participants, visualized a running system. This work investigates how the previous work can be enhanched by adding the possibility to visualize systsem behaviour at different abstraction levels and improve usability, scalabiliy and overview in the visualization. A monitor for the visulization of distributed reactiv systems will be implemented. A system model and trace information from the obeserved system will be used in order to realize the visualization. By adding the possibility to visualize nested collaborations, a systems behaviour can be visualized at different levels of abstraction. The necessary constraint on the model of a system, and an algorithm for the detection of nested collaboration in the observed system, is suggested and implemented. An automatic approach for the loading of data from a model of the observed system is added to improve uasbility. Additional filtering and layout mechanisms are implemented in order to provide better overview in the visualization. A post-mortem approach to visalization is taken in order to generate a correct visualzation. The speed of the visualzation can be controlled by the user. The possibility to reverse the visualization is also implemented.</p>

<p>This thesis describes a project that utilizes honeypots to detect worms. A detailed description of existing worm detection techniques using honeypots is given, as well as a study of existing worm propagation models. Simulations using some of these worm propagation models are also conducted. Although the results of the simulations coincide with the collected data from the actual outbreak of a network worm, they also conclude that it is difficult to produce realistic results prior to a worm outbreak. A worm detection mechanism called HoneyComb is incorporated in the honeypot setup installed at NTNU, and experiments are conducted to evaluate its effectiveness and reliability. The mechanism generated a large amount of false positives in these experiments, possibly due to an error discovered in the implementation of the detection algorithm. An architecture using honeypots for detection of unknown worms is proposed. This architecture is based on a combination of two recently published systems with the extension referred to as a Known-Attack (KA) filter. By using this filter, it is believed that the amount of traffic needed to be processed by the honeypot sensors will be considerably reduced.</p>

<p>Multicast is used to send data to many receivers simultaneously. Multicast protocols developed for wired networks are not suitable for a Mobile Ad Hoc Network (MANET), mainly because the mobile nodes create a changing topology, and the capacity of the nodes and the links are low compared to a wired network. Hence, it is a challenge to distribute information in an Ad Hoc network. A goal for this Master’s thesis work has been to investigate content distribution in mobile ad hoc networks to find possible ways to reduce necessary volume of distribution data. Scaling according to available bandwidth and alternative distribution methods to multicast has been investigated. A test architecture consisting of different components useful for content distribution has been configured and tested. Essential components have been: • Multicast OLSR (MOLSR); a multicast plugin for the MANET protocol Optimized Link State Routing Protocol (OLSR), to forward multicast data • File Delivery over Unidirectional Transport (FLUTE); a protocol that supports sending multicast data in several layers • Linux Fedora Core 5 operating system including IEEE 802.11b Wireless Local Area Network (WLAN) adapters • A topology emulator to simulate different topologies The first tests were performed using three nodes. Different bit rates and packet sizes were tested to find the best throughput. After adding two more nodes to the network it became clear that the multicast forwarding did not work properly. Many of the first tests have thus given misleading results. After finding an explanation to the multicast forwarding fault, a modification to the test architecture was done; use an older version of Linux Fedora Core. The new test results then showed that it is possible to forward multicast data using FLUTE on a MANET. As expected, the nodes receive a larger amount of the file sent as the number of hops is decreased; up to 100 percent after one hop, whereas up to 80 percent after four hops. Also, test results showed that low data rates give better throughput than high rates. The best throughput was given after resending the FLUTE session several times. The performance of ad hoc networks is less trustworthy than wired / fixed networks. The amount received varies from 0 to 100 percent. The testing of layering using FLUTE did not give any improvements. However, layered content was not available, so all the channels had the same content. Suggestions to possible ways to provide layered content are described. The concept of layering is still interesting for MANETs since it provides the ability for nodes with low capacity to receive less content than nodes with high capacity in a relatively simple way. Further tests are needed to see the results using layered content. When implementations of codec frameworks become more available, these may be great for scaling in ad hoc networks.</p>

<p>Optical packet switched Migration capable network with service Guarantees (OpMiGua), a collaboration between Telenor, Network Electronics and Norwegian University of Science and Technology (NTNU), is a hybrid network architecture which combines the principles of optical circuit and packet switching, thus providing guaranteed service in the strict sense and statistically multiplexed properties. By introducing two more QoS performance sub-classes within the statistically multiplexed property class in the OpMiGua network, further performance differentiation may be achieved. This master's thesis investigates the Access Control Policy (ACP) and Preemptive Drop Policy (PDP) QoS differentiation schemes for QoS differentiation of Statistically Multiplexed (SM) traffic in presence of Guaranteed Service (GS) traffic in an OpMiGua network. First, an overview of optical networks, OpMiGua, and Qos and QoS differentiation is provided. Next, the simulator developed is presented. Last, the performance of the QoS differentiation schemes in an OpMiGua network is quantified by simulation, compared, presented, and analyzed. The research has revealed that the GS traffic influences the QoS differentiation process of different service classes differently, and this influence is also dependent on the QoS differentiation scheme utilized for the QoS differentiation of the SM traffic.</p>

<p>Mobile phones and electronic gadgets on the market today have become more and more powerful, in both processing power and functionalities. Accessing files and documents residing at users’ home networks via a vast amount of devices is therefore anticipated. The current solution to remote home access, i.e. Virtual Private Network (VPN), is not supported on all different types of devices. In particular, limited processing power and memory footprint on mobile devices such as PDAs and cell phones are not suitable for VPN clients that require high processing power during encryption and decryption of data. In addition, setting up VPN on home network is not a straight-forward task that anyone can complete. Regular users do not have knowledge to set up and configure VPN correctly. It is obviously that we need another approach/method to access home network in an easier manner. The specified system, Mobile Home Access, will provide ubiquitous access to the home network independent of network infrastructures, platforms and terminals. This goal has many challenges and obstacles, such as firewall and routing issues as well as compatibility and the restrictions of the well-known networked file system, i.e. Common Internet File System (CIFS). Based on the previous research project, Mobile Home Access will be implemented as web services and written in Java, the platform independent programming language. XML web services offers intercommunication between applications and protocols running on various network infrastructure. Using web services, the networked file system’s services can be exposed to remote clients. The system will operate in two different modes, the reduced-mapping mode for restricted clients (mobile phones, PDA, etc) and the tunneling mode for rich clients (desktop computers, laptop, etc). This thesis will address and try to overcome challenges associated with the tunneling mode in addition to the design of the overall system.</p>

<p>Our objective is to test the performance of a real SMS Processing System. We will work with Boost Communications, a norwegian company that is developing a SMS Processing System to sell interactive SMS solutions. With our test, we want to be able to put the system performance into numbers, and to identify and eliminate possible bottlenecks. In order to achieve our objective we will use different techniques. First, we wil build an analytic model, a simple one, an we will use It to try to identtify the main weak points on the Architechture. Once built, we will run a few tests on the Architecture to identify our modeled parameters values. The second step will be to build a simulation model using the knowledge we will had achieved while doing the analityc model and the first tests. At the end, we will have to be able to point the bottlenecks and offer solutions to them. These solutions can be hardware based, or software based</p>

<p>Video on Demand (VoD) is an Internet service with a growing appeal to the mass market, and is of increasing importance to Internet service providers' revenue. This master's thesis presents a subjective assessment on the user-perceived quality of service of an imaginary VoD service. By implementing the SAMVIQ methodology of subjective video quality assessment, the state of the art video codec H.264/MPEG-4's resilience to packet loss is examined. Through the recreation of several residential usage scenarios, different amounts of packet loss is added to H.264/MPEG-4 content encoded at diversified bitrates. The results suggest that random packet loss rates above 0,1% deteriorates the perceived quality to such an extent that it is not acceptable to the end-user. High-bitrate encoded content is relatively more affected than low-bitrate content, and bursty packet loss is preferred to loss categorized as non-bursty.</p>

<p>The OpMiGua hybrid network has been proposed as viable migration approach from circuit switched all-optical networks to 3rd generation packet based optical networks. A special feature in the OpMiGua network compared with other hybrid optical network architecture proposals, is the combination of service guarantees with high performance. In order to achieve this, the OpMiGua network features two different service classes, Guaranteed Service (GS) and Best-effort service (BE). A crucial issue in this context is how the performance of these service classes, in particular the BE class, may be improved. To this aim, the Network Layer Packet Redundancy Scheme (NLPRS) is a viable candidate to reduce the packet loss rate in such networks. The NLPRS has earlier been studied in Optical Packet Switched (OPS) networks, but has never been considered used in hybrid optical network architectures.</p>

<p>In this master thesis there has been a description of what MANETs are and why they are so interesting. Because of its characteristics, the tradicional routing protocols for wired networks are not advisable for them. A specific routing protocol for MANETs is necessary. In this thesis the main groups of these protocols have been explained and some of the most commonly used of them were studied. We saw that each protocol is better in a specific environment. None of them are perfect for all the ranges of nodes mobility, traffic, number of nodes, etc. The two main groups of protocols studied are the proactive and the reactive ones. The main characteristic of the proactive is that each node maintains a route to every node in the network. Besides, it periodically updates this information. No matter if there is communication between the nodes or not. As representative examples of proactive protocols, OLSR and DSDV were described here. On the other hand, in the reactive ones the nodes only calculate the routes between those nodes that want to communicate. This kind of protocols perform in a more efficient usage of the bandwith (which is very limited in the MANETs medium) and the resources of the nodes. However, as a drawback, when the route is not available yet, the delay to achieve it can be great. The reactive protocols choosen here to be studied were AODV and DSR. In the reactive, the main problem is the delay to achieve a new route. In the proactive, it is the high usage of resources and bandwith when it is not necessary. Both, reactive and proactive also have the problem of the scalability. To solve these problems, a new kind of protocols appeared: the hybrid ones. A hybrid routing protocol combines both, the proactive and reactive to achieve better performance. The most popular of them is ZRP and its operation was described here too. None of the existing protocols are suitable for a MANET with a large number of nodes, each one of them with a different velocity and traffic. ZRP solves in part the problem of the scalabilty, but under different patterns of traffic and nodes velocity performs worse than the OLSR, DSR and AODV. Understanding the strengths and weaknesses of each protocol, a new one was proposed. The objective of this new protocol was to be suitable to MANETs with nodes moving freely, with different ranges of speed and traffic. Also, another objective was to improve the scalability of the reactive and proactive protocols. The protocol proposed here was called Penaguila. As ZRP and other hybrid routing protocols, it is based on having some nodes working in proactive mode creating areas, and comunicating this areas with other nodes working in reactive mode. The difference between Penaguila and ZRP, is that Penaguila takes into acount the speed and traffic of each node. Therefore, Penaguila tries to have each node working in the mode more suitable for itself. Also, an evaluation of the OLSR, AODV, DSR, ZRP and Penaguila has been done. Since it was not possible to program Penaguila in NS-2 because of the short time to write the thesis, it was only feasible to do a qualitative study. In this study the advantages and disadvantages of each protocol were exposed and the concluision was that Penaguila can outperform the existing protocols when: A) The network is large, since it is a hierarchical routing protocol. B) The nodes have very different speeds and amount of traffic.</p>