Abstract:
The importance of data has risen in the past decade due the evolvement of technological innovation for
user needs. Large scale of real time, rapid changing datasets are being created due to an exceptional evolution of communication networks, social networks and internet of things. The above statement shows
network analysis has become popular and important where this lead to creation of various specialized
graph systems for network analysis. This paved way for many data-parallel frameworks to incorporate
them. However, use of relational databases for network analysis is ignored though most data is still
collected and managed in relational databases.
This situation of ignoring relational databases raises a question whether relational databases have
limitation for network analytics. The relational model is inefficient for network analysis where it will take
many expensive joins to do a computation. SQL query language also doesn’t support network analysis
operations but relational databases comes with great features, such as integrity constraints, fault tolerance,
query optimization and secure transaction and so on.
This thesis presents an integrated framework for network analytics that consist of a data model that extends
support for relational databases with network analytics. This also presents a query language to manipulate
data for relational and network analytics or a combination of them. Along with that, this integrated
framework also adds a query engine that is built on top of relational database (PostgreSQL) to process queries created with the query language. The testing results prove that query engine and model introduced were able to achieve equivalent or better performance in almost all scenarios.
