Estructuras tipo sándwich

As the Policy Master moves through the selected path in the TFFST, it evaluates condi- tions and generates actions to be enforced by the executors. The executors can play the role of subject or target in the policy [24]. For example, the executor HandoverExecutor

plays the role of target, and is responsible for executing methods to evaluate conditions and perform actions. The interface of this executor is shown.

public interface NetworkingContext extends Remote {

public String getStatus() throws RemoteException; public void quit() throws RemoteException;

/*Returns the phy connectivity status of specific nic*/ public boolean isAttached(String nicID) throws

java.rmi.RemoteException;

/*Returns the link connectivity status of specific nic*/ public boolean isLinked(String nicID) throws

java.rmi.RemoteException;

/*Returns the network connectivity status of specific nic*/ public boolean isReceivedRA(String nicID) throws

java.rmi.RemoteException;

/*Sends a message to initiate selection process*/ public void networkSelectionEvent(String nicID)

throws java.rmi.RemoteException; /*Executes upward handover*/

public void executeUpwardHandover(String nicID) throws java.rmi.RemoteException;

}

There are two types of actions: internal and external. The former (e.g. networkSelec- tionEvent) are performed within PROTON. The latter, for example executeUpwardHan- dover, occurs between PROTON and the mobile host (see Figure 5.10), and these are executed by the Control Interface that lies between the network layer and the mobility management sub-layer (i.e. MIPv6 module). The interface controls the incoming router advertisements from different access networks and it executes the corresponding actions (received from the Handover Executor, according to the Networking Context and the TFFST).

After receiving packets from the network (ipv6 rcv.c routine) including RAs, these are placed in the input queue (input.c) where they wait to be processed. In the case of RAs, these are sent to the neighbour discovery process (ndisc.c) to collect the appropriate information about available routers in the network. However, in the case shown in Figure 5.10, router advertisements are intercepted by the PROTON support system and sent to the corresponding module. PROTON filters some RAs (using IPv6tables filters) and re-inserts the appropriate ones into the stack to control the handover process.

Actions are associated with the different stages of the handover process. The Control Interface runs scripts based on IPv6tables [46], as a packet filtering tool, to build the appropriate rules that inhibit automatic handovers (by filtering router advertisements) and control them according to the Networking Context. These scripts also set timers considering context (e.g., mobile host velocity) and execute the most convenient handover mechanism.

4G Communication system LAN GPRS 3G WLAN satellite 4G Communication system LAN GPRS 3G WLAN satellite ipv6_rcv.c ipv6_input.c ROUTING mipv6.c ipv6_output ROUTING ipv6_fw.c ipv6_xmit.c ndisc.c neighbour discovery router hook ads handover executor (userspace) PROTON SUPPORT CONTROL INTERFACE SYSTEM

Figure 5.10: Handover Executor implementation.

5.6

Remarks

In this Chapter, PROTON was presented, a policy-based system to support multi-mode devices. Motivation behind PROTON stems from the fact that future devices will have multi-mode capability for connecting to different wireless networks. It is demonstrated how PROTON can address several issues of future networking, and how it can cope with complexity and dynamics intrinsic to future environments.

As far as I know, PROTON is the first policy-based system that attempts to offer complete mobility support for 4G mobile users. These heterogeneous environments pose challenges that remain open. Using a policy model based on TFFSTs, PROTON helps users in many decisions while hiding the added complexities.

It is also shown that concepts from autonomic computing can be applied to the design of novel solutions that brings us closer to the answer of open networking challenges such as seamless roaming among heterogeneous technologies.

This project consolidates the idea of building the policy evaluation model on the network, to enable devices with the capability to deal with complexities while keeping a powerful light-weight solution.

The Networking Context dataset presented is rich enough to allow well-informed de- cisions while roaming. However, the possibility of using a rich set in such a dynamic environment is empowered by the idea of having three levels of information according to the dynamics of data elements.

PROTON’s architecture also reflects the concern of dealing with constraint devices, particularly in a constantly-changing environment. This is the main drive behind dividing PROTON into network- and host-side components. Every module that demands intensive computation work or high storage capacity is located in the network.

The mobile device deals, exclusively, with the evaluation of TFFSTs, a task that does not require much processing. Via the application of novel algorithms for the specification and translation of policies, conflict resolution, and TFFST distribution and installation, I have implemented a complete system that supports seamless roaming in upcoming per- vasive networking environments, while representing a light-weight solution that is easy to deploy.

Powerful and more intelligent solutions are needed to support inter-networking in fu- ture communication systems. However, mobile devices and wireless environments will always exhibit strong limitations in terms of memory capacity, processing power, and stability. Hence, the prior resolution of conflicts and TFFST distribution is a very appro- priate approach to overcome these constraints.

PROTON has demonstrated the potential of merging concepts of autonomic com- puting with the design and implementation of a policy-based system, together with a novel evaluation model and efficient conflict-resolution algorithms. The result: a solution that offers full mobility support, hides complexities, enables smart decision-making while roaming, and deals with the intrinsic constraints of 4G environments.

Chapter 6

Evaluation

The deployment of a policy-based system such as PROTON in constrained devices such as those used in mobile environments, represented a huge challenge with clear trade-offs between costs added into the network side, the link, and the end devices. Hence, a novel approach was proposed to extend the advantages of policy-based solutions beyond the wired world; it includes all the benefits of complex policy-based systems, but within a simple abstraction deployable in the wireless world.

To evaluate this work, its most relevant aspects were examined. First, using the LCE-CL testbed, a test case was simulated to demonstrate the usefulness of the system. PROTON’s main goal is to improve users’ connectivity and thus augment their produc- tivity. It is important to ensure that the system takes appropriate decisions according to its purpose. Therefore, this was evaluated during the experiment; the outcome is reported in Section 6.1.

Second, the departure from a complex policy-based system, centralised in powerful computing devices, to a simple decision-making structure, distributable to mobile ter- minals, mandates low resource usage and overheads in terms of computational power, memory, and bandwidth. These were measured during the experiments, and the results are presented in Section 6.2.

Third, Section 6.3 discusses how feasible it is to deploy PROTON in a typical mobile phone examining the main constraints: the out-degree of TFFSTs and the number of transitions (which determine the TFFSTs’ size) and the number of events.

Fourth, Section 6.4 shows the scalability of the system when increasing the three elements that bound the complexity of a policy: number of events, context fragments (conditions), and actions.

Finally, a taxonomy of mobility management systems derived from a comparison be- tween PROTON and other similar systems is presented in Section 6.5, emphasising the characteristics that lead to a good solution for mobility management. In addition, the practical advantages of PROTON are summarised at the end of this chapter.

user connected to LAN starts data transfer receives NetworkDisconnection(eth0) receives ContextChangeTransition(gps) receives NetworkConnection(hotspot) Pedestrian Mobility Profile

High Speed Mobility Profile Low Speed Mobility Profile macro-event HighAutomobileVelocity() constraint No DownwardHandover macro-event LowAutomobileVelocity() receives FaddingSignal(hotspot) receives NetworkConnection(hotspot) time (s)

sequence number _{time sequence graph}

receives

NetworkDisconnection(hotspot)

Figure 6.1: Testing PROTON in a simulated scenario.

6.1

Test case

Imagine Alice in her office using her PROTON-enabled laptop; she starts downloading a large amount of data that she needs for an important business lunch in London’s financial district. She decides to leave her office immediately and continue downloading the data on- the-move. When Alice disconnects her laptop from the local network, the first PROTON event is generated: NetworkDisconnection(eth0). This initiates the policy evaluation, the laptop connects to the Wi-Fi network available in the building and continues downloading the data without any disruptions to the connection.

When she leaves the building, the PositionSentinelcannot read the data from the in- door location system—e.g., the bat system [105]—hence generating the second event: Con- textChangeTransition(gps). The mobile device begins determining its location using its GPS receiver, and connects to an available cellular system (e.g., Vodafone’s GSM/GPRS network). As Alice approaches her car, PROTON detects a nearby hotspot and generates the event NetworkConnection(hotspot) and evaluates the appropriate set of policies to decide how to use this broadband network.

She starts driving on the highway towards the city centre, and as she drives acceleration is detected by GPS-receiver. This is monitored by PROTON that generates a macro- event: HighAutomobileVelocity()—this triggers the selection of a different mobility profile that is loaded in the device. PROTON connects to the GPRS network when hotspot connectivity is lost, and while Alice is driving on the highway no downward handovers are allowed. This is because of the constraint NoDownwardHandover specified in the

HighAutomobileVelocity mobility profile and built into the corresponding TFFST. Thus, she stays connected to the GPRS system.

She reduces speed as she reaches congested traffic areas in the city centre. This situation is detected by theVelocitySentineland the macro-eventLowAutomobileVelocity()

is sent. Another mobility profile is loaded. The NetworkConnection(hotspot) event is received: PROTON evaluates the TFFST and decides to continue with the data transfer using an available hotspot. A few minutes later, the signal from the current access point starts fading and the event FadingSignal(hotspot)is generated.

PROTON changes its attachment point without any disruption; it uses the most appropriate execution method and the optimal initiation time, and adapts itself to the new QoS conditions exploiting its policy model and Networking Context dataset. Alice arrives at her final destination, and PROTON connects to the restaurant’s hotspot to download the last few bits of data, while Alice starts her meeting.

PROTON’s usefulness was tested through simulated cases such as the example de- scribed; these show how the system, autonomously, executes appropriate actions to en- hance users’ mobile experience. For example, Figure 6.1 shows how typical network activities would be efficiently performed in a real 4G system when using PROTON.

The system was installed in a multi-mode device (a Toshiba Satellite laptop) that can access multiple access networks, and a sequence of events was produced. This se- quence triggered the evaluation of a subset of policies and generated the execution of the corresponding actions.