Google AI Professional Certificate - Learn AI Skills That Get You Hired
AI Adoption - Drive Business Value and Organizational Impact
Overview
Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Learn how to build a high-performance time-series database directly on object storage without traditional server overhead in this 42-minute conference talk. Discover how Man Group, one of the world's largest hedge funds, replaced a massive MongoDB cluster with a "thick client" architecture capable of streaming data at 40GB/s. Explore the challenges of managing hedge fund data at scale and understand why MongoDB's traditional approach created an "arms race" with users. Examine what constitutes a truly serverless database and master techniques for building atomicity and versioning on object storage systems. Analyze performance bottlenecks including CPU, I/O, and Python overhead considerations. Investigate the global state problem when managing millions of keys and dive deep into Set CRDTs, comparing grow-only versus observed-remove approaches. Understand why distributed locks can become problematic and explore future possibilities with atomic operations and conditional writes in S3. Gain insights into retrofitting metadata and compaction strategies through the Q&A session, making this essential viewing for software architects working with distributed systems, database design, and serverless architectures.
Syllabus
- The problem: Managing hedge fund data at scale
- Why MongoDB failed us The "Arms Race" with users
- What is a truly "Serverless" Database?
- Building atomicity and versioning on Object Storage
- Performance bottlenecks: CPU vs. I/O vs. Python overhead
- The Global State Problem: Managing millions of keys
- Deep Dive: Set CRDTs Grow-only vs. Observed-Remove
- Why Distributed Locks are a trap
- The Future: Atomic operations and conditional writes in S3
- Q&A: Retrofitting metadata and compaction strategies
Taught by
InfoQ