1 Overview
1.1 What is a Shard?
It is introduced in the document below
https://www.elastic.co/guide/cn/elasticsearch/guide/current/kagillion-shards.html
-
The bottom layer of a fragment is a Lucene index, which consumes a certain amount of file handles, memory, and CPU operation.
-
Every search request needs to hit every shard in the index. It is good if every shard is on a different node, but it
同一个节点上竞争
is a bit bad if multiple shards need to use the same resource. -
计算相关度
The term statistics used is yes基于分片
. If there are many shards, each with very little data will lead to very low correlation.
The relationship between clusters, nodes, and shards is introduced in the official document