| dc.description.abstract | This thesis will provide an answer to the real-time trade monitoring use case with an
implementation of the space-based architectural style and compare the performance of this
approach to the traditional three-tier architecture.
To do this, a data flow is designed to imitate the real-world trade transaction flow, and two trade
monitoring system is implemented, with the first one being a modular system using the
Hazelcast platform as the core in-memory data grid component, and the other as a three-tier
system which acts as a benchmark for the performance of the processing units and reuses the
modules where possible.
After this, a performance test suite is designed, with the main purposes are to test the main
components of each system, namely the processing units and the database change data capture
module. Using the metrics gathered, a comparative analysis method is used to testify my claim.
The test reports suggest that the three-tier architecture does not perform as well as the spacebased one even with the same configuration of computation resources. Interestingly, a
performance degradation point in the Hazelcast platform is also found when deploying multiple
instances on the same host.
To summarize, this thesis has provided a complete small-scale implementation of the spacebased architecture system that can perform many times better than its three-tier counterpart and
is more suitable for big data or heavy load processing system use cases. | en_US |
