Abstract:
Cloud computing delivers on-demand access to essential computing services
providing benefits such as reduced maintenance, lower costs, global
access, and others. One of its important and prominent services is Database
as a Service (DaaS) which includes cloud Database Management Systems
(DBMSs). Cloud DBMSs commonly adopt the key-value data model and
are called Not only SQL (NoSQL) DBMSs. These provide cloud suitable
features like scalability, flexibility and robustness, but in order to provide
these, features such as referential integrity are often sacrificed. In such
cases, referential integrity is left to be dealt with by the applications instead
of being handled by the cloud DBMSs. Thus, applications are required
to either deal with inconsistency in the data (e.g. dangling references) or
to incorporate the necessary logic to ensure that referential integrity is
maintained.
This thesis presents an Application Programming Interface (API) that
serves as a middle layer between the applications and the cloud DBMS
in order to maintain referential integrity. The API provides the necessary
Create, Read, Update and Delete (CRUD) operations to be performed
on the DBMS while ensuring that the referential integrity constraints are
satisfied. These constraints are represented as metadata and four different
approaches are provided to store it. Furthermore, the performance of these
approaches is measured with different referential integrity constraints and
evaluated upon a set of experiments in Apache Cassandra, a prominent
cloud NoSQL DBMS. The results showed significant differences between
the approaches in terms of performance. However, the final word on
which one is better depends on the application demands as each approach
presents different trade-offs.
