A new SRE function for risk
BNP Paribas set up a dedicated SRE function for its risk department, which runs critical trading, payments and risk applications on a traditional on-premises banking stack. I established the function and its practices, working with more than 50 application teams.
- Role
- Senior SRE and Transformation Lead (contract)
- Team
- Cross-functional team across production support, operations, development and architecture, plus external contractors
- Scope
- 50+ application teams · 1,000+ applications · on-premises, no cloud
The challenge
Two critical trading risk platforms were unreliable, with availability around 65%. Monitoring was reactive, incidents took too long to acknowledge and fix, and reliability practices varied from team to team.
What I did
- Defined the SRE framework (SLIs, SLOs and error budgets) and embedded it into development, staging and production workflows across the department.
- Built a proactive monitoring and troubleshooting framework for the trading risk platforms.
- Designed dashboards in Dynatrace (through OpenTelemetry), Splunk and ITRS Geneos, so teams worked from service-level metrics rather than alerts.
- Introduced tooling, workflows and automation for out-of-hours support, and mentored the offshore operations teams in SRE fundamentals.
- Led break/fix across the infrastructure, network, storage and database teams, focusing on low-latency network and storage on Red Hat Linux.
- Advised on moving critical workloads from physical and VMware servers to Kubernetes and OpenStack.
Outcome
- 65% → 99.9%availability of two critical trading risk platforms within six months
- 30 → 5 minaverage time to acknowledge and repair incidents
- −20%first- and second-line helpdesk calls
- 50+application teams adopting SRE practices
Stack
- Dynatrace
- OpenTelemetry
- Splunk
- ITRS Geneos
- Red Hat Enterprise Linux
- Java
- Python
- Sybase
- VMware
- Kubernetes
- OpenStack