Make Your Data FABulous
summary
Explains concrete accuracy vs speed vs scale trade-offs in Elasticsearch: terms aggregations use per-shard top-k (shard_size = size1.5+10 by default) which produced 60 instead of 64 counts on skewed data until shard_size was raised to 20; cardinality uses HyperLogLog++ with precision_threshold (default 3000, max 40000) and memory ≈ precision×8 bytes, and using a single shard gives exact top-k but sacrifices horizontal scale*.
AI-written summary · about
chapters (7)
- 0:00Start
- 2:03CAP Theorem
- 9:33elasticsearch
- 11:08Terms Aggregation
- 15:21Routing
- 16:14(Sub) Results Per Shard
- 28:57Single Shard
related talks
48mMake Your Data FABulous
51mMake Your Data FABulous
37mBeyond the basics with Elasticsearch
41mTips and Tricks to Scale Elasticsearch for 1M RPMs and Beyond
40mElasticSearch: Big Data, Search and Analytics
29mHow (not) to Scale Elasticsearch for Data Analytics!
24mWhen and How Should you Shard?
47mMake sense of your (big) data using Elasticsearch