ElasticSearch is a search engine technology used by many companies for use-cases like: full-text search, autocomplete suggestions, data analytics, etc.
It is a document-oriented database. The high level architecture of ElasticSearch is as follows (bottom-up):
Document
Documents are the basic units of information you can store in ElasticSearch. You store the data in the form of a JSON document. Every document has a unique ID which is used for reverse indexing the data (discussed at the end).
Index
An index is a collection of the similar type documents. An index is analogous to the database in relational databases. When you query in ElasticSearch, you query against a particular index.
Node and Cluster
An ElasticSearch cluster is a collection of nodes connected together.
Shard
It is not necessary that an index in ElasticSearch cluster need to be present only on a single node. It could be broken down into parts and shared among multiple nodes to achieve horizontal scalability. Each part of an index present on a node is called a Shard.
Example:
users index has 10k documents which are shared between node1 and node2. Each part of 5k documents is called a shard.
Now that we understand high level architecture of ElasticSearch, let's have a look at how it internally works.
Reverse Indexing
ElasticSearch uses a special data structure called reverse index. Every term in the data is indexed against the document IDs of documents containing the term.
Let's understand this through an example below:
Consider 2 documents as stated above with IDs: 1 and 2. When these documents are indexed(loaded into the index), every term from the document is taken and it is mapped to the document ID. When you query for a certain term, it is like a lookup in the dictionary. This is what primarily makes ElasticSearch very fast.
More articles to follow in this series of ElasticSearch articles. Please comment down if you want me to write an article about a specific topic.
Top comments (0)