Dynamic management of network systems.

摘要

Networks are becoming integral part of our lives. Services hosted on our mobile devices, network hubs and networked servers provide us with functionalities that were never so ubiquitous. These services are customized to our requirements more than ever and provides us with the intelligence and support that we need. Researchers are innovating cutting edge technologies and services to leverage the smart networked devices. With extensive functionalities the smart phones, tablets and other mobile and sensing devices are posing new challenges. These challenges span over various aspects i.e., from morphology of these networked devices to the mobility pattern of these devices and from policy enforcing the data shared by these devices to the data transport protocols and congestion control mechanisms.;With growing ubiquity of the mobile devices, the next generation service oriented systems face various challenges due to increasing number of services involved in each system and volume of data produced by them. The quality of the produced data or information can be sensitive to many application specific parameters and this Quality of Information (QoI) is represented by application specific utility function. Moreover, if a user is interested in getting information from a specific area then the information from or about such an area is highly relevant to the user's interest. This thesis focuses on three major challenges: (1) How to efficiently manage resources such as network routes, among multi-priority applications in resource-constrained scenarios, such as congestion, to reduce the loss of QoI; (2) how to configure and compose services that produce information which is relevant to the users interest? (3) How to share network resources efficiently by mapping human readable policies to low-level system constraints.;To address the first challenge, this thesis explores pricing mechanism for route allocation in event-driven scenarios. Multi-priority traffic suffers from loss of QoI in event-driven scenarios. Such scenarios trigger bursts of data when an event happens, e.g., sensors sensing a target. We research how to utilize pricing mechanism in routing and show that by dynamically pricing the network routes, they can be efficiently assigned to multi-priority traffic to avoid congestion thus lowering the overall QoI losses.;Sensor Networks, mobile devices and online services are producing huge amount of data these days. From plethora of such data sources, fusing data from most relevant sources to produce information that is relevant to the users interest is a challenging task. Thus, this thesis also explores how can we dynamically compose complex sensor services that produce information relevant to the users interest? We present three models to capture the concept of relevancy in composite services and present the design of a service-oriented system that uses these models to compose complex services from services hosted on sensor nodes, conforming to operational and relevancy constraints.;For better resource management and security of the network infrastructure, access to its resources is often regulated via policies based on user's credentials. Moreover, if the network infrastructure is shared among multiple organizations then resources are not always equally available to all the users. In order to regulate access to the network resources frameworks and mechanisms are needed to define, modify and enforce policies. In the last part of this thesis, we present a policy enforcement framework that enables resource-sharing among multiple parties in a shared network infrastructure. We explore how to enable end-users and policy makers to use human readable formats for system management and transparently transform human understandable policies to machine-level constraints for application configuration. We present a set language based approach to produce restricted set of resources for executable application configuration of a user request based on declared policies and user's privileges.