The customer, a leading asset brokerage, wanted to optimize its global trading platform — responsible for supporting over 70% of its equities and futures business. With business volumes expanding, a primary concern was the increased risk of end of day and overnight batch processing failing to complete on time, with only a 5 hour window to make before markets re- open for the next day’s trading.
Sumerian’s Forward Thinking predictive analytics could provide the accurate, quantified insight the customer needed into this complex process, allowing it to improve overnight batch performance and deal with increasing volumes within this tight processing window.
Using log files from the batch manager, underlying infrastructure resources and resource monitoring tool, Sumerian’s first step was to create a baseline model of the customer’s system. This provided a complete picture of system resource utilization, demand profiles for everything being processed, the performance of each batch job and the impact of each job on the duration of the overall batch.
Sumerian then analyzed the relationships between trade volumes, batch performance (both individual steps, jobs and the overall batch) and infrastructure utilization. This allowed Sumerian to understand the impact of changing business volumes on the duration of each individual batch job and the overall batch, and understand the sensitivity of each job to changing volumes. It also allowed us to understand the impact of changing volumes on underlying infrastructure capacity and identify any potential bottlenecks.
Based on this we were able to make several recommendations that would allow the customer to optimize the batch process and reduce overall batch duration. Recommendations included identifying certain jobs and tasks for parallelization or re-coding and infrastructure resources that would benefit from re-balancing.
Outcome and results
In less than 6 weeks, Sumerian Forward Thinking predictive analytics identified 10 explicit actions that the customer could take to save a total of 125 minutes in overall batch duration. 5 of these were quickly implemented and alone saved the customer over 70 minutes of batch time. This yielded an immediate 15% reduction in time taken to process the overnight batch.
As a result the customer was able to mitigate the risk of having to process increased business volumes, and successfully continued to meet the 5 hour overnight batch window.