Browsing by Author "Dobrodeev, Vladimir"

Mobile networks become more and more pervasive. The growing demand for mobile connectivity fosters installations of new base stations and various low-power network elements. Furthermore, the upcoming 5th generation of mobile networks (5G) will require to significantly increase network elements' density. The growing number of network elements introduces new challenges for network management. Among these challenges there are scalability and an ability to handle larger volumes of data, which also have higher variety and require higher processing velocity for faster reaction on abnormal situations. The requirements introduced by data volume, variety and processing velocity require Big Data solutions. Data exchange in Big Data can be organized with a special Message Oriented Middleware (MOM), which is a software solution providing a scalable and reliable message-based communication. A popular communication paradigm for MOM is publish/subscribe. There are two main approaches to design of systems based on this paradigm, namely, topic-based and content-based. The first one offers lightweight message routing but has limited capabilities to describe required content. The second one eliminates these limitations at a cost of added complexity of the message routing resulting in higher requirements to middleware performance and message processing cost. The solution proposed in the thesis utilizes benefits of these two approaches. The solution combines expressiveness in describing interests and easy message routing from topic-based approach. In the core of the solution is an algorithm realizing a perfect matching between interests and topics. Constructing such a matching offers more efficient usage of system resources, e.g., network traffic, by utilizing overlaps between interests