Data @Scale Seattle - Dharma Shukla
summary
Azure DocumentDB converts JSON documents into a unified JSON→tree model and builds a unioned path index (labels map to terms with postings) so queries run without predefined schemas or secondary indexes. The engine embeds a JavaScript runtime for UDFs and transactions, enforces consistency choices from strong to eventual, and supports provisioned throughput in request units per collection (examples: 250, 2,500). Index maintenance uses a latch‑free mapping table, blind incremental updates with 8 MB sequential flushes and commutative merges to handle bursts (thousands of writes per second) within tight CPU/IO budgets.
AI-written summary · about
chapters (14)
- 0:00Intro
- 1:07DocumentDB
- 1:50What is DocumentDB
- 4:50DocumentDB Architecture
- 8:07The Problem
- 11:42Indexing
- 14:29Querying
- 15:56Query Il
- 16:58Logical Representation
- 18:05System Model
- 19:15Challenges
- 20:32Classical Btree
- 23:08Log Structured Storage Organization
- 26:41Semantic Merges
related talks
53mManaging enterprise JSON workloads with Amazon DocumentDB
33mDriver for Data-Driven Change - Azure Cosmos DB Performance, Security & Cost Optimization Done Right
01h 02mNot your Father's SQL Server: Developer Innovations in Data Storage
52mData storage for modern applications - Mike Lehan
54mSay Yes to NoSQL for the .NET SQL Developer
25mData modeling, the secret sauce of building & managing a large scale data warehouse
45mJSON - a first class citizen in Azure SQL DB
01h 02mZero to Document Hero