5 boring performance improvements in MongoDB 3.4 that matter
summary
MongoDB 3.4 speeds real tasks: initial sync builds indexes in parallel so a ~660GB test drops from about a week to under 4 hours; balancer moves off mongos to the primary config server (single authoritative balancer); wire-level Snappy compression cuts WAN replication traffic and can make cross‑continent syncs 4–5x faster; default removes migration throttling for WiredTiger and allows up to ~N/2 concurrent chunk migrations (one thread per shard) so large-scale rebalancing completes far sooner.
AI-written summary · about
chapters (9)
- 0:00Start
- 4:30Faster Initial Sink
- 20:13Thread per Shard
- 22:14Balancer Time Window
- 29:16Parallel Chunk Migrations
- 31:57Secondary Reads in a Shattered Cluster
- 46:04Steam Compression
- 47:00Disk Encryption
- 50:15Hardware on the Config Server
related talks
20mMongoDB size management
28mComparing MongoRocks, WiredTiger and mmapv1 for performance and efficiency
40mData Migration Lessons from the Trenches
45mSharding and Things We'd Like To See Improved
39mTake Your Percona Server for MongoDB to the Next Level
25mWhat’s New in Spring Data MongoDB
39mMongoDB vs MongoDB: How, When and Where to Use It
47mSchema design for fast applications