A modern, scalable risk system architecture
summary
A clear architecture that treats every datum as versioned immutable keys and pushes heavy pricing to instrument-level jobs rather than position-level aggregation, reducing data movement and making reruns exact. Prototype maps cleanly to cloud: SQS + DynamoDB with Python and boto3, workers read small job specs (instrument ID, timestamps, scenario) and write results back; controller only coordinates. Notes real-world scale limits (historical deployment reached 600 workers and hit data‑center capacity), and position scaling is done offline.
AI-written summary · about
chapters (20)
- 0:00Intro
- 0:16What is a risk system?
- 4:59A historical perspective
- 6:11the humble spreadsheet
- 6:43Humble Spreadsheets
- 8:05Use a Database
- 8:55Version 2.1 - Client /Server
- 11:49Version 2.3 - Heath-Robinson Spreadsheets
- 12:48Heath-Robinson in Practice
- 14:51Distributed Systems
- 18:06Calculation Loop
- 22:12The Future
- 23:13Core functions of a Risk System
- 25:30Everything Changes
- 31:26Versioned Datasource
- 35:47Trade Processing
- 39:34Position Aggregation
- 45:37Calc Loop
- 49:26Design Principles
- 49:57Overview
related talks
23mAn Open Source Architecture for Modernizing Market Risk Management
37mShaping the future of modular and composable banking solutions
37mReactive Architecture on AWS: from 0 to 1.6M events/minute
51mImmutable data stores and CORS for safety, flexibility and profit
43mSoftware Quality As It Relates To Data
59mBuilding resilient event-driven architectures
36mThe MODERN Modern Data Stack: Building an Open Distributed Data Warehouse Beyond...
41mEnhancing Event Analysis at Scale: Leveraging Tracking Data in Sports