Nfq has evolved the original economic capital tool of one of the largest spanish retail bank, developed in Matlab, which ran once per year and spent 12 hours, to a Spark approach running every month and spending 40 minutes.
Nfq has evolved the original economic capital tool of one of the largest spanish retail bank, developed in Matlab, running once per year and taking 12 hours, to a Spark approach running every month and taking only 40 minutes.
The client has a group’s economic capital tool, developed in Matlab, which takes too long to run.
Due to time constraints, they are forced to reduce the number of executions during the year and the characteristics of each process (number of scenarios).
The client had acquired a new infrastructure, based on Oracle Big Data Appliance and Cloudera, on which the further development would be based on.
Nfq has optimized the process from a methodological and informational point of view.
Nfq has redesigned the new tool focussing on the parallelization techniques provided by Spark.
Nfq has provided a PaaS infrastructure during the construction and testing phases due to the delay of the hardware delivery.
Nfq has deployed the tool in the new infrastructure helping with the optimization during the configuration process.
Benefits for the client
- Reduced dramatically the execution time.
- Improved the reporting and analysis capabilities.
- Brought new calculation parameters to the user and closing the process to the business strategy objectives.
- Explored the best practices and limits of the new infrastructure.
Problems to solve
- Need of increasing the number of simulations
- Change to a new infrastructure
- Need of pre-deal calculation
Where we put our focus
- Process Optimization
- Data Analytics and Methodology
- Oracle Big Data Appliance
- PaaS infrastructure on Amazon Web Services
- Hadoop ecosystem: Spark