School of Engineering & Design

Chapter 1: Introduction

Autonomic wireless access systems was a model created in order to support the development of communication networks in the direction of efficient flexibility and extensibility to a large sort of potential faults and attacks. On the other hand, particular importance is given on the foundation values to accomplish focused behaviour over a self-organization. Therefore, an autonomic system allows the performance of an autonomic network when including self-management, self-configuration, self-optimization, self-healing and self-protection and the relation with different numerous of dynamic network groups and communities. As a result, every day more people, private customers and enterprises are employing wireless access technologies in order to achieve better connection, mobility and easier reception at any location. However, as the significance of this wireless technology has been increasing with the time, the need of autonomic self-management systems became crucial. In fact, this dissertation will collaborate to the actual investigations related with autonomic wireless management systems. In addition, when considering autonomic networks is critical the examination of protection requirements and the administration of security infrastructure. The system presented in this report will illustrate a qualitative evaluation and simulation results that shows self-healing characteristics and autonomous conduct. It presents some of the disadvantages found when trying to simulate an autonomic behaviour and the restrictions of using software simulators.

1.1Towards Self-Management.

Self-management [1] is the core of autonomic networks, and its intention is to liberate systems administrators from the facts of system operation and maintenance and to give users with a machine that can be used at all times. In addition autonomic systems will preserve and regulate their operation in the case of changing components, hardware failures, software malfunctions, and outdoor conditions. The autonomic system has the capacity of frequently supervise itself, and check for component improvements.

Self-configuration takes part when the system can configure itself in harmony with high-level procedures. In the case of a new component is adapted, this will incorporate itself easily, and the rest of the system will adjust to its incidence.

Figure 1.0: Features of self-management [1].

Self-optimization means the way that several parameters must be set correctly for the network to work ideally. Autonomic systems try to improve the operation of different technologies by identifying and gathering their main functions in order to make them much more competent in performance or cost. Therefore, is here when autonomic systems will monitor and experiment with their individual factors in order to make proper choices, verifying, and applying the most recent updates.

Self-healing takes place when the system can identify, trace and determine the origin and cause of a particular malfunction in complex mobile ad hoc networks [2]. On the other hand, serious user problems can occur and their explanation can take several weeks to be detected and fixed. However, autonomic systems will discover, analyse, and restore localized failures by using information about its individual system configuration and may be through a regression tester.

Self-protection over autonomic systems can be applied in two ways. Firstly, we can protect the network as a whole against considerable, associated problems coming up from attacks or consecutive failures. Secondly, we can predict problems based on premature reports obtained from sensors and take steps in order to evade or make them smaller.

Table 1.0: Four aspects of self-management [1]

In mobile ad hoc networks, it is very common to deal with unexpected alterations of population, topology and the complex reconfiguration executed by end-users. In addition, it is highly difficult for people to manage and to have the control of these unpredictable changes because the complex character of mobile ad hoc environments due to the heterogeneity of components and technologies, the lack of infrastructure and the decentralized composition.

The model to follow in this project about mobile autonomic networks is based on a specific type of wireless mobile ad hoc networks [3]. In other words the network is built spontaneously as devices connect, instead of relying on a unique central base station to manage the current of messages to each node in the network, the individual network nodes forward packets to and from each other.

A mobile autonomic network could be a semi-centralized, mobile ad hoc, wireless, autonomic network, where its nodes are heterogeneous and the whole network can be easily manage by non-expert users. Therefore, a kind of mobile autonomic network can be for example, a home wireless network, a SOHO wireless network, an emergency service ad hoc network a business meeting network or a military scheme ad hoc network.

The security needs that the mobile autonomic networks required are mainly the same as those usual networks. In other words, the security architecture obtained from wireless, mobile and ad hoc networks might be implemented in mobile autonomic networks.

1.2 Problem Depiction

Self-managed wireless network posses a quantity of characteristics that allow us the developing and creation of autonomic base stations or access routers. However, in order to achieve the best performance of the network there are some problems that must be taken into account.

The number of capabilities that every base station must enclose includes:

• A base station may be self-configure as long as it has a home duplicate of running and administration information in addition with particular installation of the management methods.
• Self-optimisation characteristic includes the facility of a base station to communicate with other ones and to find their positions on the network. As a result, each base station must be connected each other since they have to share important information during the authentication procedure.
• The security part needs special attention since all the interchange of data among base stations has to be protected against attackers in addition with the reliability of the management information such base station exchange (self-protection).
• Self-adaptive attributes must take place when every base station inform to the others about any modifications in its environment.
• It is essential for the network to have a self-managed communication method that guarantee a well performance as the traffic and the interchange of information increases. The idea is that every station will be able to operate and interact with others by creating new configurations or, be capable of running its default configuration sequence.
• In the event of failures o errors every base station must be capable of find the solution and solve the problem itself or, if necessary, notify to other access points in order to correct the failure globally as a method of self-healing conduct.

1.3 Aim and Objectives

To present and simulate an autonomic behaviour through the use of self-healing capabilities and the improvement of some of the different aspects involved within a decentralized management system when being applied to an autonomic wireless network context in addition with the analysis of all requirements that an autonomic base station has to contain.

• Describe a decentralised concept for a group of autonomic base stations.
• Illustrate the manner how self-protection, self-healing and self-optimisation are applied to autonomic wireless networks.
• Evaluate the possible solutions when improving the scalability problems of autonomic management systems.
• Simulate an Autonomic behaviour using OPNET Modeller by showing how self-healing capabilities may be adopted.
• Analyse some of the different drawbacks presented when trying to evaluate autonomic conditions.

1.4 Dissertation Organisation

Chapter 2 looks into what are the existing approaches for autonomic systems and the management of wireless networks. Additionally, it examines some of the issues when dealing with autonomic networks and essential network technologies involved in the autonomic networks field. Chapter 3 describes the proposed decentralized management system and different autonomic applications may possibly run on it. Furthermore, important security characteristics will be discussed. Chapter 4 presents the modelling and simulation, using OPNET Modeller 14.5-Education version for the autonomic wireless network management. This chapter contains all modifications and suppositions that were necessary in order to achieve an autonomic self-healing behaviour including failure detection and results. Finally, conclusions, simulation´s limitations and discussion of further work in Chapter 5.

Chapter 2: Literature review

2.1 Autonomic Systems

Autonomic systems can be denoted as interactive groups of autonomic elements, individual systems components that have resources and deliver services to specific users and other autonomic elements. These elements may manage their inner actions and their associations with other autonomic elements in harmony with policies that people or other elements have recognized. In addition, a spread, service-oriented infrastructure will maintain autonomic elements and their communications.

Figure 2.0: Structure of an autonomic element [1].

Figure 2.0 shows an autonomic element that consists of one or more managed elements attached through a particular autonomic manager that controls them.

The managed component will be alike to what is found in ordinary non-autonomic systems, although it can be personalized to allow the autonomic manager to observe and manage it. The directed element may be a hardware supply, such as a CPU, a printer, or a software resource, such as a database [1].

Within a more complex field, a managed element may be an application service, electronic business application or other related IT application. The autonomic manager makes a distinction the autonomic element from its non-autonomic equivalent. If we can monitor the managed element and its exterior atmosphere, and constructing a plan based on an examination of this information, the autonomic manager will reduce people of the dependability of directly managing the managed element. The autonomic computing is expected to progress as designers slowly add progressively more complicated autonomic managers to existing managed elements.

Each autonomic element will be responsible for managing its own interior condition and conduct and for controlling its relations with an environment that consists mainly of signals and communication from other elements and the external world. An element´s internal behaviour and its relationships with other elements will be determined by objectives that its maker has implanted in it, by other elements that have authority over it [2].

The element may need help from other elements to accomplish its goals. In this case, it will be in charge for obtaining essential resources from other elements and for dealing with exception cases, as an example the failure of a necessary resource.

Autonomic elements will take part at many levels, from individual computing components such as disk drivers to small-scale computing systems such as workstations or servers to entire automated enterprises in the largest autonomic system of all the global economy.

At the lower levels, an autonomic element’s variety of interior behaviours and relations with other elements, and the set of elements with which it can act together, may be comparatively partial and hard-coded.

Mainly at the rank of individual components, well-established techniques, many of which fall under the rubric of fault tolerance, have led to the expansion of elements that not often fail, which is one significant feature of being autonomic.

When having higher levels, fixed behaviours, links, and associations will give way to enlarged vitality and flexibility. All these aspects of autonomic elements will be articulated in more high-level, goal-oriented terms, leaving the elements themselves with the responsibility for resolving the details on the fly.

Service-oriented architectural models such as Web and network services will play an elementary function, a sufficient foundation for autonomic computing requires more. Primary, as service providers, autonomic elements will not obediently honour requests for service, as would classic Web services or objects in an object-oriented environment [2]. In fact, autonomic elements will provide only services that can improve the reliability of its objectives. Autonomic elements will initially a problem requests to other elements to carry out their objectives.

After all, autonomic elements will have complex life cycles, frequently transporting on various threads of activity, and constantly sensing and responding to the environment in which they are situated. Autonomy, proactively, and goal-directed interactivity with their environment are distinguishing characteristics of software agents.

2.2 Architecture of Autonomic Networks

The current consideration on network organization remains. For example, autonomic communications holds the layer as:

• User
• Applications
• Services / Components
• Middleware
• Network and Computers
• Device
• Agent (soft) or Sensor (hard)

Take into account that this layering has network, computers and software, transmissions, and information. But the requirement of supporting everywhere networks is required; a strong domain model must be incorporated in Autonomic Communication architectures [3].

Domains substitute layers as the main organizing component. Domains will supply limits, much like the border and doorway routers do today in IP networks. But domains will need to mechanically communicate large amounts of information about what they will and will not accept; what they are capable of, and what present facility and QoS is accessible. Domains will contain domains, each with alike computational manage systems but with diverse locations, contexts and communities [4]. Criticize domain structures will have connection and provide shortcuts across these many networks.

A distributed computing, tools and platforms will be a fixed part of the network. While routers and switches will still pass data, they will subcontract all but the simplest and quickest of decisions to the implanted distributed computing substrate; only things like lookups will stay.

In result, every mechanism will have a service interface with available methods. In some architectural systems, the device-embedded service will dynamically register itself with service modellers and discovery components – principal of which is the agent-service providing self restoration. This probably means that every device will include Java or .NET and provide service and runninginterfaces accessible by soft services.

Basic to the plan of all autonomic networks is a fixed security model. In conventional networks, protection was applied after the reality - mainly via external devices, applications and tools. The world would be a much more organized and safer place if TCP/IP had been designed with fixed security and even this is being looked at again [5,6].

Autonomic networks must be self configuring, self deploying, and self assembling. While total security cannot be assured and security does add fee, for the most part of current designs have produced workable compromises where the security of a network is well known. The best designs can adjust cost and security as properties of network domains, controlled by policy and bounded on the borders with a security that filters user and service admittance.

Figure 2.1: Four basic steps are designed into Autonomic Network [9]

Four basic steps are designed into Autonomic Network [7]:

• Collect: get together information from environmental and device sensors and instrumentations; determine purpose or service requirements; look up user context and security credentials.

• Analyze: this function uses the current tools to associate and model multifaceted situations for example, queuing models, economic models, rules and policies. In addition, these mechanisms permit the autonomic manager to use the network environment more efficiently and to predict further conditions.

• Decide: process with the best available computation engines including decision theory; risk analysis, hypothesis generation, genetic algorithms, and neural networks; scope actions to situations and needs.

• Act: call up services; collect components; manage association; configure managed elements; signal devices; inform users and administrators; log actions and the strategic analysis that determined them.

2.3 Existing Management Solutions

With the use of centralized, distributed or hybrid solutions it is possible the management of wireless networks. Centralized systems make use of a unique master mechanism to configure the base stations. On the other hand, decentralized, distributed configurations do not use that single point of malfunction and collaboratively execute a completely distributed administration solution. Within any of the approaches mention above, the objective is that a wireless base station must have a reliable, system-wide composition.

Figure 2.2: Centralized topology [10]

Nowadays, with the use of wireless switches is possible to connect base stations and make them operate as wireless bridges to a switched wired network as a solution that many companies have been created in order to provide centralized administration solutions for groups of base stations [5]. With the use of wireless switches within a centralized architecture is possible to obtain channel management in addition with bandwidth, access and traffic control therefore the link-layer switch executes the management component. However, this type of centralized link-layer topologies also has disadvantages. There are several problems related with broadcast traffic that does not allow the free develop of link-layer broadcast domains. Furthermore, the architecture of the wired system does not allow straight association of the management system to the base stations. However, there are some solutions that work at the network-layer resolving this limitation.

One way to configure mobile ad hoc networks (MANETs) [16] is through decentralized management systems where its centre of attention is to enable peer-to-peer communication within very high dynamic, mobile structures [4].

Figure 2.3: Decentralized topology [11]

It means that each base station does not make any decision based on a central manager but it does base on its local capacity. There are many uncompleted researches that try to design self-configuring solutions for MANETs. On the other hand, this report is focused mainly on the configuration of fixed wireless networks for mobile clients seeking the way to improve effectiveness and performance. In addition, the important factor here is to develop an autonomic and decentralized management platform that can sustain different sort of management functions.

Hybrid advances to wireless network management give some more functionality from a fundamental system into the base stations. However, hybrid systems which are a little more complex than the plain wireless bridges do not totally attend to solve the disadvantages of centralized systems; e.g., being central points of failure for one.

2.4 A Quantity of Problems of Autonomic Home Networks

With the introduction of the self-characteristics (self-management, self-organization and self-configuration [12]) some issues have been brought within a dynamic environment. The IP-based networks have been increasing its complexity rapidly which affect directly to network designers, network operators and subsequently network users. Actual IP networking technologies execute spontaneous adjustment by responding to changes in the location once a problem has happened. Therefore, is the job of autonomic networks find the way to develop the capacity of networks, terminals and devices to deal with the unexpected variations, along with the physical and logical features of the network that can be accessed but also, changes in topology, load and other responsibilities. In addition, by the introduction of self-characteristics it provides users with the advantage to focus on their duties instead of configuring and managing networks.

Nevertheless, within a dynamic environment a quantity of problems can also be brought by utilization of the self-characteristics, such as be deficient in learning capacity of the policy-based [12], autonomic control-loop, particularly in the mixed and mobile context in addition with the complex security and cost-effectiveness of the inherent monitoring methods found in the current autonomic networking technologies.

There are some factors that must be considered when analysing the self decision makings process within autonomic networks, such as context information for routing decision which could limit the autonomic ability of the systems. Therefore, in order to be able to analyse and consequently understand the modelling and performance of the most recent autonomic networking technologies and systems, we must analyse some of their issues in the context of some basic network technologies, intrinsic monitoring systems and mobility management.

2.5 Essential Networking Technologies

The inclusion of autonomic features into modern internet architecture let us visualized a future internet system having capacities such as self-optimizing, self-management, self-organization and so on. Nowadays autonomic networking technologies are yet inadequate to adjust to diversity of upcoming network situations as follow:

Control/Decision plane: at this part the modern models of control loop just provides autonomic tasks like the policy-based autonomic control loop [13] has no learning features, providing the network with no autonomic capabilities. In addition, for an end-to-end performance goal makes so much difficult for a decision element obtain the correct decision.

Better comprehensive context technologies: in order to make decisions, autonomic network nodes obtain essential information through contexts and for this reason network contexts achievement technologies have obtained special treatments. On the other hand, contents of applications and service characteristics are very significant context information for autonomic nodes when optimizing network routines and quality of service of users.

A clear example may be, since wireless networks have a limited data transmission capacity, packet loss rate is pretty high. In the case that, autonomic nodes choose to drop a number of packets, users will suppose that important packets are going to be protected while insignificant ones to be dropped. Therefore, with a better comprehensive context aware technologies will provide such behaviour by gaining and analyzing the context data of clients, applications and network.

An obedient network: total responsiveness regarding to objectives, businesses and applications must be essential properties of Self-Managed Autonomic Networks.

2.5.1 Monitoring Infrastructures

Actual autonomic systems are facing problems regarding the way monitoring infrastructures provide local and global performance information consistently and efficiently not including security problems and performance degradations. In addition, the status concerning the autonomic network must be sensed by an autonomic system which leads to a large amount of data flow to be handed out to the autonomic system.

Therefore, the monitoring infrastructures must be strongly built with very high performance, the security of the principal monitoring system have to be cautiously calculated as well as the expenses of big monitoring activities ought to be take into account. Additionally, in case of have a large size monitoring activities further processing and combining into the total performance information for the autonomic systems and services will be required due to a massive production of data.

2.5.2 Mobility Administration

Mobility management becomes much more complex and complicated since potential mixed wired and wireless systems are going to supply everywhere coverage and flawless mobility in order to allow additional users with multi-mode transportable terminals. As a result, for a mobile device will be a problem quick to response different changes perfectly and opportune and acquire suitable calculations without human intervention along with heterogeneous access situations. Likewise, in order to accomplish competent resource employment, load contribution, bandwidth aggregation and some more features, mobility administration must support interface-level [13], mobility and flow-level mobility.

The junction rate of the control-loop in the autonomic networks might be low since diverse types of the heterogeneous access technologies, the time-changing characteristic of wireless channels produce vibrations when collecting information for the control-loops. As a consequence, it may have effects on the self-configuration and self-adaptation process in the mixed wired/wireless situations.

Chapter 3: Design of an Autonomic Management System

3.1 Overview

The aim of this chapter is to describe the design of the autonomic management system. In addition, it will explain what sort of suppositions must be taken into account when choosing the autonomic access points as well as the most important security characteristics and the basic network’s functionality. After all, the chapter will describe some important autonomic functions that can be including in the central system.

3.2 General Requirements

The model that is presented in this report requires of a well selected base stations which are fundamental components of the autonomic wireless system. IP [22] access routers can be used as wireless base stations due to their assigned IP subnet and they have the capability to correct any error that may occur without the intervention of any human operator.

Each access point will require a wireless interface in order to provide wireless services to every user or mobile piece of equipment connected. Another network interface which can be wired or wireless is going to supply uplink [25] communication to mobile devices and the other part of the network. In fact, it can be used for Internet access [14]. In addition, this second interface may be use as a supervision and management interface among base stations and to support self-management services.

The use of supplementary interconnection lines can be useful when connecting mobile devices that come from different networks (wired or wireless) and there is a need to link different channels and protocols.

The description mentioned above is represented in the Figure 3.0 where four access routers provide wireless interfaces for connectivity and uplink among different mobile devices. The other network interface is used to transfer data between the access router and the Internet and for autonomic administration capabilities. However, an extra or extras uplink interfaces could be included.

Figure 3.0: Access Routers with their wireless interfaces (uplink), administration interfaces and additional interfaces.

It is the job of the access routers to allocate and configure IP addresses and subnets which are going to be used for the mobile devices and for the uplink interfaces.

Self-protection (Figure 3.1) can be achieved by following a number of considerations that the model proposed in this report must make use of. This model will use X.509 certificates [15] along with public-key cryptography in order to obtain key distribution and user authentication. A key pair (private/public) is created by a certification authority (CA) [17]. Public keys are included into a certificate; they do not need to be protected from unauthorized nodes and may be spread with the use of software [24], information servers or an Internet system.

Figure 3.1: Self-protection using X.509 certificates.

A two-way authentication along with the use of X.509 certificates provides reliability in many security aspects. Every base station needs to corroborate its neighbour’s certificates. Consequently, each base station must contain the related certification authority´s public key and the device´s certificate. At the same time, the use of hash algorithms is fundamental in this application, being SHA-1 [18] and MD5 [23] the most commonly known and used. The Secure Hash Algorithm (SHA) generates a 160 bit (20 byte) message digest in contrast with the 128 bit (16 byte) message used by MD5. Therefore, SHA-1 is stronger against attacks but slower than MD5.

3.3 Functionality of the Network

An essential idea about this report and the projected model is to show how we can distribute the administration functionality of the wireless system, in some way that no central point of failure will exist and as a result will be avoiding bottlenecks [18] during the processing and communication of information. For that reason, all the base stations must support autonomic management capabilities and be able to provide self-protection, self-healing and self-optimisation. In addition, important information regarding the wireless network configuration and running activities must be exchanged and collected among each base station and its neighbours hence every station within the radio range work together as a uniform wireless access network.

Within a network with inner management station, there is communication between each station and the central administration station but no exchange of information is possible between other base stations. Avoiding central point of failure [14, 18] topologies, interchange of irrelevant information among neighboured base stations is evaded. Hence, the information recovered and collected locally will permit base station to choose the correct management configuration.

The information exchanged between base stations sometimes is selected, based on the significance of the information for example, if it can be spread to neighbours or be kept private. Exchange data is just passed to a single access router and it must be absolutely reliable all over the network. In addition, this information will be extend globally among stations which are closely situated and every station have to report to its neighbours regarding any modifications by using sporadic periods in order to avoid massive transit of messages.

Figure 3.2: information exchanged between next-door base stations.

Figure 3.2 represents how information is shared between stations and the use of next-door stations. In other words, when a base station obtains new management information, this data will be transmitted to the nearest neighbour. The station which received new information must inform to its next-door station about any internal modifications including previous data received from the initial transmitter [20]. However, if a base station wants to start management information modifications, the data used will not be forward aging because to the initial sender this information is not new any longer.

The autonomic configuration occurs when access points receive and exchange certain information among other devices and used for specific purposes. Every base station posses an autonomic behaviour since they shared information each other but none access point can be obligate to make use of information or incorrect configuration procedures. In addition, this protects all devices within the wireless network against attackers that want to manipulate base stations and force them to execute malicious procedures.

3.4 Categorisation Process

It is very functional to make a segmentation based on the type of information that every base station manage in order to make easy for the base station to decide what sort of information it should used during a specific process.

3.4.1 Global Information.

Each base station decides how the information is handled, based on the importance of this one. Global information refers to all sort of information that requires be globally distributed to every collaborating base stations. An example of global information may be an access control list (ACL), which is a record of mobile nodes that are authorized to join the wireless access network.

Current base stations a number of software packages must be installed in order to make available its service, for example, a DHCP server, the operating system, a firewall and the administration software itself. Global information includes a list with actual versions of software; together with the mode they can be retrieved. Furthermore, unknown mobile nodes or base stations also called rogue stations could bring several risk factors into the managed wireless access network for example; deny service to legitimate users, start man-in-the-middle attack and attempt against wireless performance. Therefore, global information may include a list of stations that are not authorized to connect to any base station in the network in order to reduce the risk of attacks (self-protecting mechanism).

3.4.2 Public Information

Public Information will be exchange only among next-door devices which can be found within the radio coverage. As a result, this information contains the number of devices connected to each base station, the connection path use for base stations and information regarding addressing. Consequently, when a group of next-door stations want to communicate each other public information shared allows them to act in response to events happening within a specific area.

By using bi-directional broadcasting at regular intervals to next-door devices, base stations use its public data to revise and renew information regarding to its home status as well as its own system updating.

Figure 3.3: Style of information that every base station manage.

3.4.3 Private Information

Private data will be found and used exclusively for local proposes and it is not transferable among devices. Security related information and capacity data will not be shared with anyone. However, some security elements are exchanged throughout the entire network.

3.5 Performance and Setting up

Propose of this section is to explain how the suggested autonomic wireless access network works based on its management functionalities. Firstly, we are going to analyse the performance of the self-configuration method, the course of action when adding new base station. Afterwards, we will be discussing how to include homogeneity when processing global information and after all explain how a centralized configuration could be transformed and used within a non-centralized autonomic wireless access network.

3.6 Improving the Scalability of the Network

The process of initialization of a base station starts establishing its synchronisation [20] follow by a checking period. Subsequently, the base station may start configuring itself in order to supply connection mode to all mobile devices associated to each node of the network. In case of power malfunction, all stations may be re-initialized themselves at the same time, but because they are not synchronized they will not be looking for new next-door stations in a parallel way.

After a small period of time where the base station generates a probe of connectivity this is ready for an auto-configuration IP process, which includes designation of subnets in order to bring association capabilities to mobile devices and get ready with the uplink line [20]. In addition, the base station looks for its next-door devices with propose of testing their correct configuration and the later inclusion within the self-management network.

At this part of the process beacons [26] are imperative elements because they provide to a neighbour base station with network management information as a result of the quantity of overhead that the broadcast of beacon frames produce which is very significant; conversely, a beacon is capable of identify the presence of an access point. By using a radio network interface card (NIC), every radio frequency channel is scanned looking for beacons approaching from base stations in order to discover an appropriate one.

Figure 3.4: Beacon Frame (Guaranteed Time Slot (GTS)) [26].

As soon as a beacon is discovered, information regarding groups of access points and potential information about the network could be obtained. Therefore, association process among bases stations may start taking place.

Power reduction mode is another important attribute of beacons. Inside our self-managed wireless access network, a base station periodically transmits mapping information using beacons in order to identify which base stations utilizing power reduction mode have data frames claiming for them at the access point buffering service.

In case that none next-door stations can be found by the access point during the searching process, the device autonomously will use a default configuration and then be able to commence operation among mobile devices.

Figure 3.5: Wireless scanning process looking for next-door base stations.

Figure 3.5 illustrates the manner how an access point initiates a wireless scanning process looking for next-door base stations. As soon as a station detects the presence of the other access points within its coverage area the station ratifies the scanned access point as one of its neighbours.

Resolution of management interface addresses takes place with the transmission of a resolve request by the base station to its next-door previously scanned stations. The resolve request may be sent as a user datagram protocol (UDP) [21] message. As a result, broadcasting and multicasting messages can be used. On the other hand, if fresh stations use encryption or filtering the communication may possibly fail.

A new access point gets association with all the addresses and stations found throughout the wireless scan using their management interfaces. This is like the MAC address of the interface which makes possible wireless services.

In order to obtain association with next-door stations the recent base station may utilize the wireless interface as a mobile point of connection at the same time transmitting the solve request to that base station. Therefore, needless multicast or broadcast messages could be avoided inside the wireless access network. However, a fresh base station could broadcast or multicast the resolve request by using its administration interfaces.

If the neighbour stations do not provide answer at all to the new base station, it will set up the configuration by default as well as connection service among mobile devices. The first time neighboured stations receive information regarding a new base station is when this one sends to them its resolve request. A successful authentication of a new base station is conceded by its next-door stations, only they will authorize this procedure using the information exchange during the presentation part and the implementation of an X.509 certificate previously discussed in the section 3.2.

Transmission control protocol (TCP) [22] guarantees reliable delivery of management information exchange among base stations by establishing a full duplex virtual connection between devices. At this instant, the new base station is capable of provide all its neighbours with public and global information.

The new base station will be ready then to combine all the data recovered from its different next-door stations during the searching phase and the selectivity of available radio channels. In addition, a base station performs a series of scans in order to identify alterations around its surroundings. However, the scanning procedure could affect user’s connectivity depending on how frequently is executed as a result regulation take place during the uplink transmission among mobile nodes.

3.7 Network Database and Recovery

Several advantages will be obtained with the use of databases that save and store important information regarding all devices inside the wireless access network. In other words, a network administrator could be capable to obtain significant data about the actual network configuration and status just by using previous collect related information.

In case of network failure, the fact of having a backup of all information related will be incredibly helpful in addition with the security fact that none malicious attacker will be able to modify or utilize certain information because it will be out of reach.

One way to apply this concept is with the creation of virtual neighbours [25]. In fact, a neighbour will be adding to a list of all actual next-door stations no matter if it belongs to the same radio coverage. In this order of ideas, the virtual device will collect the same amount of messages that direct or (real) neighbours have received.

Figure 3.6: Database and transmission of management information using a virtual next-door device.

Scalability is easily obtained with the inclusion of virtual devices since a station does not differentiate among a virtual or real next-door device. Therefore, a virtual neighbour simply recovers the configuration information of a pre-selected subgroup of each and every one stations contained by the wireless access network.

3.8 Security Requirements and Technologies

The use of security applications must be imperative when trying to guard the wireless access network. The system has to be protected against malicious infiltrations, illegal accessing, interception or modification of private data [24]. In addition, authentication information and encryption keys have to be protected from harm.

Masquerading attack [19] is quiet difficult to detect and one of the most dangerous attacks, it can affect the normal operation among mobile clients and base stations. An invalid entity may impersonate a valid base station and acquire illegal control over the traffic network.

Figure 3.7: Masquerading attack. Itaffects the regular operation between mobile clients and base stations.

The management procedures carrying out within the wireless network could be used as a way to attack the network. A broadcast including resolve messages is created by the enemy using a station’s address as destination address. In consequence, the station affected will receive several answers due to the broadcast attack sent to all active base stations.

Data integrity and privacy among base stations is possible by using the transport layer security (TLS) [8] protocol. The TLS protocol is divided into two layers:

• TLS record protocol
• TLS handshake protocol

TLS record protocol uses private connection and encryption is achieved using symmetric cryptography. The connection is trustworthy and message transference makes use of a message integrity check using MAC address. SHA-1 and MD5 can be used as hash functions.

TLS handshake protocol provides authentication, encryption algorithm negotiation and cryptography key checking between stations.

Chapter 4: Simulation and Results

4.1 Overview

The aim of this chapter is to illustrate the modelling and simulation, using OPNET Modeller 14.5-Education version for the autonomic wireless network management. In addition, it will explain what kind of modifications and suppositions were necessary in order to achieve the autonomic self-healing mechanism, including agent´s architecture and description.

4.2 Autonomic Management Agents

This section will illustrate the modelling and simulation, using OPNET Modeller 14.5-Education version, of a community of autonomic management agents that provide network fault analysis for a group of base stations. The main objective of these intelligent agents will be to bring together process information in order to detect failures when bases stations exchange information between them and the creation of a high obtainable wireless access network. Analysing network failures is relatively difficult since these problems may differ from one network system to another and could depend on network dynamics, i.e., the type of network information to be exchange and the traffic characteristics associated with that information. In addition, the pattern of failures could vary quickly as the network operates and reconfigures around a failed device.

As OPNET Modeller 14.5-Education version does not have autonomous process ready for simulation usage, existing code had to be adapted to allow autonomic behaviour. The use of two different autonomic agents we required in order to provide self-healing network diagnosis and facilities. In this report, OPNET coding modifications will be called Agents and two different types are mentioned and applied to the access points.

Testing Agents will supply data simplicity and monitoring capabilities to Node Agents, on the other hand, Node Agents will check periodically the information that Testing Agents bring together and use it as a medium of failure detection within the wireless access network. In addition, a Testing Agent will be able to supervise and provide data regarding information exchanged among access points. Node agents use data obtained by the Testing Agents as a method of node analysis.

Various Testing Agents may be found on a single wireless client. A Testing Agent can be situated on a host device since it does not have to deal with data acquisition and information simplicity. In contrast, Node Agents will be located on a base station. Various Testing Agents may be found on a single wireless client.

4.2.1 Additions and Model Modifications

OPNET Modeller was used in order to determine concept achievability of the proposed model. The conception about Autonomic Mobile Wireless Networks is illustrated by using a community of wireless base stations which allow autonomous healing of interrupted paths. Intended how autonomic healing (self-healing) is possible in broken routes.

The OPNET simulation showed in this report will contain two Node Agents and two Testing Agents which take part of a group of autonomic base stations. The new OPNET topology required the creation of ten nodes in order to characterize every autonomic agent and all the modifications were made to accomplish the needs of both agents. The autonomic behaviour was obtained through modifications to the wlan_server_adv and ip_arp_v4 OPNET process models, where code changes were made in order to achieve the desired behaviour.

Figure 4.0: OPNET ip_arp_v4 process model.

4.2.2 Testing Agent (TA) and Node Agent (NA) Description

Each Testing Agents belong respectively with a Node Agent as a single component of a particular node in the OPNET simulation. As mentioned in section 3.3, each base station is aware of its next-door stations at all times. A Testing agent (TA_1) is designed to watch and detect alterations regarding other base stations. In the event of any modification of the network, TA_1 will notify Node Agent (NA_1) by using a UDP message. UDP presents lack of reliability consequently; the Testing agent TA_1 cannot assured successful message transmission. However, this lack of reliability will be useful for simulation proposes.

After receiving information from TA_1, a Node Agent (NA_1) will inform other stations about changes in zone, and files updating may take place. When a NA_1 observes that information sent has no arrived to its destination by a particular period of time, the agent alerts its neighbours that a probable node malfunction had happened. This time depends of certain attributes fixed for a particular mobile user.

Scalability of the network will be achieved with the use of a second pair of agents. Where, Agent TA_2 has the job of monitoring path request messages sent and received by other stations. Information regarding path request is detected by TA_2, including the time when the path request was generated and the destination of this demand.

Changes to the mobility architecture were necessary including ARP and IP alterations. The idea was to alter some settings in order to evaluate and compare the destination address with the address of the device were specific information was sent. The destination address must belong to a registered wireless client and the intelligent agents will check correct transmission of it.

IP alterations were made changing the moip_core to allow stations to be able of forwarding information packets to its neighbours, modifying the IP routing mode and helping each station choosing the better route available. The moip_core has a list that could be dynamically regulated as the base stations travel between networks.

The UDP is used as a transport protocol and the managing, mobility and registration information is handling by the process show in the figure below.

Figure 4.1: OPNET moip_reg process model.

The moip_reg process allows base stations when managing and updating mobility information regarding next-door stations. When exchanging information among stations, all the agents will monitor and process each request and they will aim to find failures during the registration process. If the registration communication was successful, there is an identification value that is compare with a mobility list and the right matched among them will mean no error has occurred during the registration procedure.

Updated messages must be sending when agents have no information regarding the mobile station due to updating failures. In fact, agents need acknowledgments in order to be sure that the communication between stations is doing perfectly, in case that an agent does not receive the updating message, it will not be able to monitor base stations and all the information exchanged among agents will be lost. Therefore, all the updates and acknowledgments will be verified within an identification field contained by the moip_reg. If they are equivalent, the update will be set as confirmed and the exchange of information will be free of failures.

Figure 4.2: OPNET agent node structure.

Figure 4.2 shows a plain representation of agent node structure and distribution. In addition, OPNET Modeller allows us to present the node model which was modified in order to provide autonomic behaviour to a set of autonomic base stations within a self-managed wireless access network. The wireless connectivity is achievable through the use of IEEE802.11b interfaces, permitting roaming among networks. This type of interfaces could be improved by adding an extra communication module between the radio transceiver and the wlan_mac system. This process allows a base station to simulate the effect of completely losing connection among devices and at the same time avoiding unnecessarily queues of packets.

4.3 Network Model

Three different network configurations were constructed to simulate and identify autonomic characteristics, and agent distribution was arbitrarily decided in order to improve the simulation. The Testing agent (TA_1) was applied to a single base station, other station was selected to make use of Testing agent (TA_2) and Node Agent (NA_1) while Node Agent (NA_2) were modified to operate in all base stations.

4.3.1 Design of Wireless Network Infrastructure

The next steps were followed in order to design a wireless infrastructure in OPNET:

1. Open the OPNET program and select NewProject and the press OK.
2. Give the project a name and the scenario name.
3. Select create empty scenario and press next.
4. Network space was chose as campus and specific size was selected as: X-span and Y-span 10 kilometres respectively.
5. The Object Palette Tree will open which illustrates the various WLAN devices as follows.

Figure 4.3: OPNET Wireless-LAN-Adv.

The file Node Models situated in the object palette contains the item wireless-lan-adv which encloses all the different network devices used in wireless network presented in this report.

All nodes were modified by using the function configuration Application Config and Advance Edit Attributes option.

Figure 4.4: OPNET Advance Edit Attributes.

In addition, the following wireless parameters were customized as stated in the figure 4.4:

• Physical characteristics
• Data Rate (bps)
• Transmit Power (W)
• AP Beacon Interval (sec)
• Packet Reception-Power Threshold (bytes)

The wireless access network contains ten base stations (Figure 4.5) which are connected via point to point duplex links (ppp_adv). Each base station has at least two interfaces; one interface to provide connectivity among wireless mobile devices and another wired interface in for uplink communication.

Figure 4.5: OPNET Network Model 1 and Agent Node representation.

The network configuration showed above was created in order to simulate and analyse the wireless system when it includes nodes (Base Stations) on the exterior sector of the network and no more than two neighbour stations close to it. Therefore, these stations will only have a maximum of two paths to communicate their next-door devices. On the other hand, the rest of base stations will be surrounded by more stations and more possible routes.

Figure 4.6: OPNET Network Model 2 and Agent Node representation.

Figure 4.6 shows the second configuration. There are various potential routes on which base stations and mobile devices may exchange information among them. As a result, agent’s performance is going to be probe by selecting the best path and being able to repair route problems.

Figure 4.7: OPNET Network Model 3 and Agent Node representation.

The third model illustrated in the Figure 4.7 offers a more narrowly linked network configuration. The number of neighbours for every node will increase and the communication between Node Agents and Testing Agents will improve due to a decrement in the number of paths required for Testing Agent information to meet the suitable Node Agent. Therefore, a superior self-healing performance will be expected using this configuration.

4.4 Verification of Agent’s self-healing process upon base station malfunction.

To experiment the right operation of the agents, different simulations were made in every network model. The main purpose is to test agent reliability and its competence when providing intelligent self-healing course of action. Consequently, the base stations were programmed to reproduce a failure and the action of agents would eventually lead us to simulate an autonomic behaviour.

Figure 4.8: OPNET Autonomic behaviour, Self-healing approach.

In order to obtain a more understandable vision of the self-healing performance, a reduced network configuration was simulated (Figure 4.8). Exchange of information among nodes may take different paths until data arrives at its final destination. In the event that a particular base station fails, the permanent monitoring service of the Node Agents will detect the malfunction, and then base station’s self-healing method will autonomously locate another route allowing intelligent diagnosing and repairing.

OPNET code modifications provide one method of simulating a malfunction in the base station. The most important features required for this process was the use of acknowledge mechanism and understanding of the range capacity of base stations. These characteristics were required to allow mobile devices to recognize when a failure takes place in a base station and stop transmitting and routing traffic, in order to start self-healing and path recovery.

4.5 OPNET Simulation Results

A set of simulations were run in order to obtain some important information regarding self-healing and path recovery. The figure 4.9 shows the interchange of data being sent by the two mobile devices and obtained by mobile_node_2.

Figure 4.9: OPNET traffic among mobile devices.

The red line indicates the data and traffic being sent by mobile nodes 0 and 1, while the blue one represents information received by the mobile_node_2. In order to detect the presence of devices and the best possible path, control traffic and management features are established and can be noticed by looking at those short spikes at the beginning of the simulation. In addition, there is a stable amount of traffic being transmitted without interruptions.

Figure 4.10 below illustrates the how the information is transferred by using the two base stations. No information is sent or received from base station 2 due to its disabled condition.

Figure 4.10: OPNET information is transferred.

It means that Node_2 is routing the same amount of traffic accepted by the mobile nodes.

Figure 4.11 indicates the End-To-End delay (≈0.0165s) of information from the mobile nodes. There is a preliminary connection between mobile_node_1 and the base station. As a result, the simulation shows that mobile_node_1 has a reduced amount of delay comparing to mobile_node_0.

Figure 4.11: OPNET end-to-end delay among mobile nodes.

The following scenarios will show the performance of the network when using self-healing conduct. The simulation results will facilitate to comprehend the intervention of autonomic agents and its autonomous behaviour (self-healing).

Figure 4.12 shows an autonomic behaviour of self-healing. The green line represents a spike at the five minute of simulated time; this spike is caused by the self-healing option of the autonomic agents, when the autonomous base station realizes a route failure it device performs route finding in order to locate the most favourable path for the data being transmitted. Therefore, one wireless router might fail but other router will be able to pick up the information and brings it to its destination.

Figure 4.12: OPNET traffic among mobile devices, Self-healing approach.

The failure action can be observed at the five minute of simulated time; the light blue line shows self-healing performance and represents a new base station receiving traffic from all next-door devices which provides new path for communication. Finally, the blue line overlapping with the red one indicates the traffic sent and received by the mobile nodes.

Figures 4.13 and 4.14 show the information exchange between the mobile nodes and the node_2 (see Figure 4.8). The amount of data being received by node_2 is represented by the blue line. On the other hand, the red and green lines show the traffic being sent by the base stations to node_2.

Figure 4.13: OPNET Autonomic behaviour, Self-healing approach.

In this node, the agent performance constantly permits the exchange of information with an individual occasion of break at the five minute simulated time. When the self-healing path detection takes place the autonomic agent try to find new routes and provides management and traffic control to a final destination.

Figure 4.14: OPNET WLAN Data traffic, Self-healing approach.

Figure 4.15 indicates the End-To-End delay of information from the mobile nodes to the final destination. There is a small amount of information being dropped due to router failure. Approximately at the five minute simulated a small break indicates data being dropped. However, the end-to-end delay is stable during the whole simulated period which means high level of reliability retained in network exchange information even with a device malfunction.

Figure 4.15: OPNET end-to-end delay among mobile nodes, Self-healing approach.

4.6 Failure Detection and Results.

For straightforwardness, a set of different simulations were run in order to check the reliability and performance of the autonomic agents and the self-healing behaviour. To verify if the agents could identify a node malfunction, a set of recreations were implemented for each of the four network topologies and some attributes were added to each base station to create the failure condition. Therefore, the results obtained will depend of the harmony among agents and their capabilities when detecting and correcting communication problems. Each agent will inform every failure identified directly to the mobile device and the results of each simulation will be put into a table.

The different tables presented below, showed the results of the simulations. The rows indicate when a node failed in detecting the problem or when the agent effectively identified the failure. However, a base station that contains a Node Agent previously set to fail could not be responsible if the failure is not detected.

4.6.1 Network Configuration 1

The table below indicates the result of different simulation runs, the effectiveness and behaviour of the Node Agents. Both agents were able to identify some of the total probable errors separately. In most failure cases both agents detected the same error which means that exist great harmony within the system. On the other hand, only one node failure was totally unnoticed.

	Node Agent NA_1	Node Agent NA_2
Base Station 1	Error	Failure Detected
Node_2 (NA_2)	Failure Detected	Error
Base Station 2	Failure Detected	Error
Base Station 3	Failure Detected	Error
Base Station 4 (NA_1)	Failure Detected	Failure Detected
Base Station 5	Error	Error
Base Station 6	Failure Detected	Failure Detected
Base Station 7	Failure Detected	Failure Detected
Base Station 8	Error	Failure Detected
Base Station 9	Failure Detected	Error

Table 4.0: Failure Detection, Network Configuration 1.

4.6.2 Network Configuration 2

Both agents completed the simulations with good results. Using Node Agent NA_1 seven failures were detected. On the other hand, Node Agent NA_2 discovered only 5 errors but still great harmony in the network system. On the other hand, only one node failure was not detected by any of the agents.

	Node Agent NA_1	Node Agent NA_2
Node_0	Error	Error
Node_1	Failure Detected	Failure Detected
Node_2	Failure Detected	Error
Node_3 (NA_2)	Failure Detected	Failure Detected
Node_4	Failure Detected	Failure Detected
Node_5	Error	Failure Detected
Node_6	Failure Detected	Failure Detected
Node_7 (NA_1)	Failure Detected	Failure Detected
Node_8	Failure Detected	Failure Detected
Node_9	Failure Detected	Error

Table 4.1: Failure Detection, Network Configuration 2.

4.6.3 Network Configuration 3

Among all the set of simulations performed, the last simulation we run was the most successful in terms of failure detection since no failure was totally undetected. Each Node Agent was capable of detect 9 of 10 possible node failures. In fact, no malfunction was absolutely unnoticed, which provides a superior harmony level and self-healing capability.

	Node Agent NA_1	Node Agent NA_2
Node_0	Failure Detected	Failure Detected
Node_1 (NA_1)	Failure Detected	Failure Detected
Node_2	Failure Detected	Failure Detected
Node_3 (NA_2)	Failure Detected	Failure Detected
Node_4	Failure Detected	Failure Detected
Node_5	Error	Failure Detected
Node_6	Failure Detected	Failure Detected
Node_7	Failure Detected	Failure Detected
Node_8	Failure Detected	Failure Detected
Node_9	Failure Detected	Error

Table 4.2: Failure Detection, Network Configuration 3.

·New Route Discovery

The new route discovery was obtained through modifications to the wlan_server_advand ip_arp_v4 OPNET process model, where code changes were made in order to achieve the desired autonomic behaviour. In a wireless access network, if the base station and mobile nodes are within transmission range of each other, an ARP request can be use in order to find a new route to the target mobile node. The Internet’s Address Resolution Protocol dynamically translates IP addresses to its MAC level address.

When a new route is discovered, the information needs to be sent to an explicit destination (Mobile node) as specified in the “Destination Address” attribute. If the destination address specified is correct generate a destination and forward the appl_packet to the MAC layer with that information (see code below).

if (destination_address == OMSC_AA_AUTO_ASSIGN)

{

curr_dest_addr = OMSC_AA_AUTO_ASSIGN;

oms_aa_dest_addr_get_core (oms_aa_handle, &integer_mac_address, (int) mac_address);

curr_dest_addr = integer_mac_address;

}

else

{

//Destination Address" attribute.

curr_dest_addr = destination_address;

}

// Set this information in the interface control / information to be sent to the MAC layer

op_ici_attr_set_int64 (wlan_mac_req_iciptr, "dest_addr", curr_dest_addr);

// Install the control information and send it to the MAC layer

op_ici_install (wlan_mac_req_iciptr);

op_pk_send (pkptr, outstrm_to_mac);

send_paket = op_ici_create_fmt ("appl_packet");

sendID = (SPkt *) op_prg_mem_alloc ( sizeof (SPkt) );

}

In order to make the code modifications a simple as possible, the new path discovery was made through a simple Request-Response communication between base station and mobile node. Transmission of selected configuration parameters from the base station to the mobile node is possible by the creation of the autonomic agents and the interaction among agent TA1 and TA2 is illustrate in the Figure A.

Figure A: Interaction and packet exchange using TA_1 and TA_2 configuration.

The agent’s configuration is also executed in the OPNET Modeller Simulation by the use of the Node Editor described in the Figure B.

Figure B: TA_1 and TA_2 Node Models.

Chapter 5: Conclusions and Future Work

This document proposed an autonomic solution along with the required components in order to set up an autonomic configuration while providing self-characteristics to a non-centralized and self-administrative group of autonomic base stations taking part of a wireless access network. We showed how to avoid centralized topologies including the central point of failure issue. A qualitative evaluation and simulation results allow us to demonstrate self-healing characteristic and autonomous behaviour within the system.

· This report presented and simulated an autonomic behaviour through the use of self-healing capabilities and the upgrading of different aspects involved within a decentralized management system when being applied to an autonomic wireless network environment. A fast reaction of the self-managed wireless access network is very important in order to react to failures, errors and interferences. Therefore, the self-healing behaviour used in this report offers a way to address such connection problems, including a vision of how a base station can react to changes in its environment and react on failures of its neighbours. On the other hand, the proposed system is capable to handle various base stations in wireless access networks through diverse topologies.

• This management system for a wireless access network was able to identify and to adapt alterations in the network and, to be efficient, establish the alterations on all regarding base stations. The self-healing behaviour is protecting network information being exchanged and the traffic characteristics associated with that information. As a result, avoiding that the intensity of failures could vary quickly as the network operates and reconfigures around a failed device.
• The self-management of the system shows how exchange of information between nodes can take diverse routes until data arrives at its final destination, when a particular base station failed, the permanent monitoring service of the Node Agents detected the malfunction, and then base station’s self-healing method autonomously located another route allowing intelligent diagnosing and repairing.

· Alterations to the OPNET mobility architecture showed one method of simulating a base station malfunctions. In addition, with the use of an acknowledge mechanism and understanding of the range capacity of base stations the system could recognized when a failure took place in a base station, stopping transmitted and routed traffic, and beginning self-healing and optimal path recovery.

• Simulation of autonomic characteristics including self-healing, self-characteristics (self-management, self-organization and self-configuration) and security features in OPNET Modeller 14.5-Education version is a problem for various reasons. OPNET modeller requires enormous alterations of coding programming in order to achieve autonomic behaviour. Therefore, self-characteristics are not easy to apply and gather together in the same network environment because OPNET Modeller 14.5-Education version does not include autonomous process ready for simulation usage. As an alternative, modifications of OPNET architecture have to be done in order to accomplish such propose.
• OPNET Modeller 14.5-Education version does not support WAP (Wireless Application Protocol) security characteristics. Therefore, simulation of security features within wireless environments is quite complex.
• Simulation results seem to be trouble-free but the cause for this performance is that during the experiment the mobile nodes did not move due to OPNET limitations and they gave simply information from their stationary position. Further studies should involve mobility of mobile nodes and the integration of external information including a reduced amount of manual configuration using a human administrator.

5.1 Future Work

It is very important to think about future applications but more importantly is the upgrading of this actual one. The integration of external information will improve the scalability problems and the lack of self-management characteristics. In addition, the inclusion of more management applications is vital; the evaluation of base stations in wireless access networks with superior topologies is needed in order to provide autonomic characteristics such as, self-protection and self-adaptive.

More OPNET architecture modifications must be include accomplishing further autonomic capabilities, including implementation with greater number of nodes and base stations. Furthermore, OPNET architecture can be modified by the use of algorithms which allow experiments and simulations using larger geographic areas.

Cite this page