A fortune 500 bank needed to modernize their existing SAS Grid Infrastructure in order to accommodate the requirements of an upcoming new accounting program. Current expected credit losses (CECL). The client needed faster execution of the financial models in order to facilitate decision making.
In order to meet the performance requirements and create a more scalable system for the future, Stratacent recommended a set of infrastructure and design changes to adapt to the upcoming CECL program. After analyzing the current system’s workload and additional computing requirements, Stratacent recommended an upgrade to the system to include both computing power and storage.
Stratacent reviewed the current load and the performance of the existing system. Our team remodeled the new workload that will be added to the system with the upcoming CECL program and determined how performance will be impacted. The team then evaluated the possibility of upgrading the existing environment by adding more compute and storage and discovered the existing system could not be scaled due to the existing limitations. The existing system architecture could not be scaled up due to existing limitations.
We created a separate SAS GRID to handle the CECL Program load. The new design used IBM Spectrum scale (GPFS) instead of Redhat clustering (GFS2). For optimal performance and high availability, the new SAS Grid architecture leveraged physical servers, flash storage, and load-balanced batch / mid-tier web servers, and metadata servers.
After deploying the new grid to run the CECL model, the execution time was reduced from 5+ days to less than 2 days, resulting in frequent runs of the model. Further code optimization is planned to run the model even faster.