Expanding Fineract capabilities, a practical example
summary
Client growth from 11k to 430k clients and 12k to 500k savings accounts drives frequent connection timeouts and out of memory crashes under 40–100 tx/min. Fixes include swapping Drizzle for mysql connector, replacing Tomcat pool with HikariCP, moving to Azure OpenJDK and JVM tuning, disabling slow query logging, disabling backdating of transactions, changing code to avoid loading full transaction histories, splitting Fineract into three instances, multithreaded interest posting and async reports to S3 — result: heavy jobs cut from hours to minutes and single transactions from ~30s to ~300–400ms.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:14Sharing my screen
- 2:40Agenda
- 4:04About me
- 5:04Brief history
- 8:55Challenges
- 17:10Slow Transaction Processing
- 24:06Fineract Architecture
- 26:34Job Processing
- 30:51Recommendations
- 35:14Audience Question
related talks
54mHigh Performance Financial Service Integration
10mBuilding High-Performance Financial Data Pipelines
22mBlack Friday Brilliance: Managing a Billion Transactions with Tech, and Teamwork
38mMaintaining highly distributed services with Erlang: 20y experiences in fintech
46mBlack Friday Brilliance: Managing a Billion Transactions with Tech, Tactics
13mOptimizing Backend API Performance
34mHow to scale a Ruby webservice
45mMigrating Mission Critical Workloads from Mainframe to Open...