Abstract:
Maintaining the applications under optimum performance conditions are crucial. Out of many
components which affect the performance of a system, the backend databases which involved
in data persistence have been known to be affecting the performance by a high marginal.
Identity and Access Management Systems (IAM) maintain large amounts of heterogeneous
sensitive data in backend databases and it was identified there is an impact of the database
performance for the overall identity system performance. Therefore the research would be
based on performance optimization on the backend database of IAM systems.
There are mainly two approaches to database optimization as physical design optimization and
configuration based optimization. While both the approaches have been analyzed, the project
is developed around configuration parameter optimization. The author has researched problems
faced by database administrators while trying to optimize databases using configuration
parameters. To overcome the problems identified, an automatic performance optimization
framework was proposed. The solution targets supporting the database administrators with time
and resource-consuming process of optimization. ISTuneUP provides the flexibility to choose
the performance metrics which is needed to be optimized while the process could be carried
out as single-objective optimization or multi-objective optimization. Ability to set up the
configuration boundaries based on the resources allocated on the system is an added advantage
for the administrator. ISTuneUP incorporated a dashboard to provide an intuitive visualization
of the details of the performance testing. Optimization process is carried out as an Offline
tuning on the back-up database as the setup needs constant server restarts which cannot be
carried out in a deployed system. But the database administrator has given the ability to set the
identified optimums to the deployed system once the process has been completed.
ISTuneUP was able to provide optimum configurations within a considerable amount of time
with a low computational cost. With excellent potential on the project, future improvements
were identified which can lead the research to a more generic optimization framework. The
current architectural design provides the flexibility to incorporate the identified future
enhancements without harming any existing internals components.
