|
The aim of this project is two fold: First develop the monitoring and measurement infrastructure that will collect relevant data from the Bell network and data sources and feed the metrics tree; second focus on the applications of the metrics tree in addressing specific planning and traffic engineering questions. The latter will require the identification of new business and customer levels metrics and their integration in the metrics tree model. The expected result of this project is to develop a software tool, we call Cyberplanner, which will use the metrics tree model (see Fig. 1) to provide on line advice to strategic planners and network engineers. Several applications can benefit from Cyberplanner including traffic re-engineering decision support, service design and evaluation, network protocol design and evaluation, and business policy rearrangement.
The CyberPlanner software will incorporate networks, services, customers and businesses into a coherent system. This system can be utilized in analyzing the effects that changes to underlying networks, e.g., adding a new router, might have on the higher layers. Similarly, the effects of adding new customers and new services can be analyzed as well. In addition, CyberPlanner can be used to decide the actions that are needed at the network level to accommodate new business goals.
Specifically, CyberPlanner encompasses the following research activity components:
Activity 1: Business, Customer and Service layer metrics and their correlations
The objective of this research activity is to identify a series of metrics in order to characterize the performance at the customer, service and business levels. For each such metric, its precise definition, its maximum/minimum/default values, and its methodologies in acquiring the value of the metric will be provided. The acquisition of the performance data is just the starting point. Afterwards, the acquired data has to be analyzed and correlated into a coherent view so that the network performance can be revealed at a glance. The coherent view in this research activity is provided thanks to the extension of the metrics tree model to encompass the higher layer metrics (business and customer layers). The metrics tree model can be instantiated against any partial or complete real-world network. By supplying the values for the leaf metrics (Activity 2), the values of each non-leaf metrics can be computed from their children up to the business level.
Activity 2: Measurements and data collection
A Peer-to-Peer measurement infrastructure will be established to provide fine-grained, timely and precise measurements of networks under investigation. The measurement infrastructure will be able to accept measurement tasks, locate necessary measurement facilities on the network, and fulfill the task subsequently. Being a Peer-to-Peer system, the measurement infrastructure will be capable of supporting a large number of measurement nodes, thus providing a coverage that existing measurement facilities cannot compete with. In addition, the measurement infrastructure will be free from problems such as denial of service attacks and central point of failure that have plagued Client/Server based systems.
Activity 3: Optimization and simulation of network resource provisioning and capacity planning decisions
The ultimate goal of this project is to provide an effective framework to shed light into network healthiness, to identify network performance issues efficiently and effectively and, more importantly, to provide a decision support system for planning network evolution. This research activity will focus on incorporating traffic engineering and capacity planning concepts into the metrics tree model in a way that will allow deciding whether a traffic re-engineering is necessary and whether an upgrade is mandatory and profitable. In the later case, an optimal traffic distribution scheme will be derived and an optimal upgrade plan will be given. Subsequently, the performance of the traffic distribution scheme and the upgrade plan will be acquired through simulation in a non-disruptive manner. Using the simulation data, the metrics tree is recomputed and the effects of these changes can be analyzed.
|
