Abstract for Petre Dini and Andrei Negulescu
|
|
|
|
Abstract for Petre Dini and Andrei Negulescu
|
End-to-end QoS Establishment using Mobile Agents
Petre Dini, Andrei Negulescu Computer Research Institute of Montreal dini@crim.ca, anegules@crim.ca
Mobile agents is a uniform paradigm which encompasses synchrony and asynchrony, message-passing and object-passing, and stationary objects and mobile objects. Remote programming has an important advantage over procedure calling both in performance and customization. Large classes of applications can be deserved by this promising technology :
- electronic commerce in distributed systems ; - mobile tests for network and systems components ; - end-to-end QoS (quality of service), QoS negotiation, QoS evaluation, resource reservations ; - feature interactions during the service creation and service deployment ; - Internet and intranet collaborative applications.
This project deals with the end-to-end quality of service (QoS) establishment using the paradigm of mobile agents. Several issues of QoS based routing algorithms for supporting resource reservation and QoS satisfaction are supported by many negotiation and reservation protocols. Currently, the QoS routing problem could only consider a single QoS parameter, usually the number of hopes or the average path delay. However, routing under multiple QoS constraints raises NP-complex problems. For solving these aspects we have proposed a server-based negotiation and experimented specific negotiation protocols.
We try to experiment the use of mobile agents to establish the end-to-end QoS. Mobile agents will be instructed with service-oriented policies to locally adapt the performance of system components involved within a connection and/or reservation. To support multiple QoS parameters, such agents carry addition QoS information concerning the user profile, as well as available bandwidth for each link between visited components within a set of paths in order to calculate the QoS bounds. The final decision could be taken in various ways :
- changing the end-client profile with the client agreement ; - adapting locally the end-servers and intermediate components within the selected paths by considering some end-client given relaxation, or - ensuring a graceful degradation with an appropriate service cost.
The mobile agents will be empowered with QoS pricing mechanisms to select one of the most appropriate solution to satisfy end-to-end QoS.
Mobile Agents Controlling Telecommunications Services
Andrei Negulescu, Petre Dini Computer Research Institute Montreal (CRIM) {anegules, dini}@crim.ca
Building services using new development methods
One of the most important issue of nowadays telecommunications networks is the quickly and uncontrolled development of end-user applications and related network services implementation, that potentially create network instability and could seriously alter the specific network QoS parameters. Observing which services will be frequently used become critical to appropriately sizing the telecommunications networks. Observing the end-to-end performance experienced by end users of a particular service is another critical input to properly operate a network. There are also many possible interactions, that we are not aware of, between services already implemented. Though they are developed following the same methods, all these services interactions manifest in different ways, depending of instantaneous conditions and the network environment in which they exist. We need to carefully analyze the services behavior, which could no more be done from a network central point. A new dimension should be added to the existing network processing environment : mobility. The mobile ‘observers’ should also adapt their behavior, given the mobile environment in which they will be used. The services control methodology should allow the prediction of the performance of a service, and also give us the necessary knowledge on how services should be designed to improve their performance and control the demands on network resources. Because of the numerous problems and the lack of experience about new services development in a new technology environment, service providers will require very effective mechanisms that protect the network from devastating congestion conditions and users from focused service using congestion. A solution could be the observing and services control using mobile agents.
New Telecommunications Services Development and Operation Using Mobile Agents
Regarding the telecommunications services cluster, we can synthesize the following sources : · wireless networks : personal communications services, personal and terminal mobility ; · broadband transport and access networks : multimedia services (tele-education, tele-medicine etc.), data transport service; · virtual private network services. In order to handle the more and more greater network resources demand without dramatically modify the telecommunications networks infrastructure, we propose the control of the services operational state, which will no longer be acting as closed systems. We can offer this way a solution for handling the service interaction problem, another major obstacle for the achieving a rapid service provisioning. For each new service, in the service development cycle, we model parameters as the service operational state (enabled or disabled), number of service users, network resources allocated (bandwidth class, communications buffer size) and location address (updated in switch initialization phase) .
The mobile agent should be designed according to the service parameters specification. Upon service malfunction alarm receiving, the network management center application will decide to send a mobile agent which will evaluate the local switch context. Based on the results obtained after performing the implemented set of tests, the agent begin to execute the available operations (disable some services execution or reset some services parameters). The presence of a mobile agent on the switch side is necessary for multiple reasons, such as : · assuring the user transparent services ‘time division multiplexing’, that is a close control service availability on the user side ; · keep track of quickly services parameters and network resources availability changes ; · observe services behaviors and detect the eventual existing services interactions ; · observe and update customer profiles which describe preferred services, features and the delivery schedule ; · retrieve network parameters, filter and process them before sending to the management application. presentation to the end user. The use of mobile agents for this kind of applications will help to better understand the potential implications of these new services, and to characterize their behavior in a real environment. It will also allow to predict the impact on QoS, and develop flexible algorithms capable to make the best trade-off among the competing network resources demands. Of course there are some additional issues, such the agent authentication. The agent should register and the switch should be ensured that the operations performed by the agent are sufficiently secure. Another important implementation issue is that the network management center application must be able to continuously track the mobile agent location.