Performance tuning and benchmarks

Agenda

The Document describes various parameters which can be used to improve performance of XCRO applications

Deal creation

With a typical setup of EXECUTION pod scaled to 10 instances, DEAL pod scaled to 5 instances and BFF pod scaled to 3 instances, we can expect the following numbers:

Deal execution

With a typical setup of EXECUTION pod scaled to 10 instances, DEAL pod scaled to 5 instances and BFF pod scaled to 3 instances with a tolerance window of 20 minutes, we can expect the following numbers where the value in column N implies the number of executions scheduled at the exact same time e.g. BOD 08:00 AM. These executions are spread across a certain number of deals, indicated in the second column. Note that executions at a deal level happen sequentially due to prioritization, but deals itself execute in parallel. This implies that one deal having 100 executions, versus 100 deals having 1 execution each, will have different performance results.

Key assumptions

1. The results below indicate a round trip of 200ms to 300ms between the XCRO and the IL hitting CBS. These results will vary from one installation to another depending on the latency introduced by the CBS to respond to account balance inquiry and payment API calls.

2. That the CBS is capable of handling concurrent hits in the range of 20-25 per second. XCRO does not throttle and aims at hitting the CBS as quickly as possible.

Report generation

With a typical setup of --max-old-space-size=4096 for DMS pod, we can expect the following numbers to generate reports.