Make Your Data FABulous

Your Elasticsearch query returns the top 10 results. But what if the real top result is missing entirely? Here's why.

#1about 7 minutes

Understanding the CAP theorem for distributed systems

The CAP theorem states that a distributed data store can only provide two of three guarantees: consistency, availability, and partition tolerance.

#2about 3 minutes

Introducing the FAB theory for datastore tradeoffs

The FAB theory proposes another set of tradeoffs for data stores, where you can only pick two of three attributes: fast, accurate, or big.

#3about 7 minutes

How terms aggregation trades accuracy for speed

Elasticsearch's terms aggregation may return inaccurate counts by default because each shard only considers its top local results to improve performance.

#4about 8 minutes

Inconsistent relevance scores in distributed full-text search

Full-text search relevance scores using TF-IDF can be inconsistent because inverse document frequency is calculated per-shard, not globally.

#5about 2 minutes

Using a single shard to ensure data accuracy

Forcing an index to use a single shard guarantees accurate aggregations and relevance scores by eliminating distributed calculations, but sacrifices horizontal scaling.

#6about 1 minute

Why you must consciously choose your data tradeoffs

It is crucial to understand and explicitly choose the tradeoffs in your data systems, like those in the CAP and FAB theorems, to avoid unexpected behavior.