As we have seen in a previous post, we communicate with the ElasticSearch server via messages that are sent to a server. On the other hand, the server responds in messages that are received by the client. This system of messages are labelled as s “RESTful” structure. This RESTful structure is based om messages that contain all information necessary.
Such message from a client is composed of 3 elements:
- The host that is addressed
- The action that is required from the host
- All information necessary to be processed by the host
This information is stored in one message only. This message is complete: it is known by the system which node needs to do what with all necessary information. As an example, let us see how data are stored in ElasticSearch. Within the message, we know which node governs the ElasticSearch database; we know what needs to be done (PUT indicates data must be stored) and with the accompanying json file, se know the data that must be stored.
The server responds with a message that contains the result of the action, whether the action was successful or not along with additional information if required by the client.
The advantages of this structure is that it provides a stateless situation: the client messages do not depend on other client messages. This allows client messages to be handled in an arbitrary sequence.
This also allows decoupling: if the server decides to change its programme or if it is decided to distribute the programme over different nodes, no change in the messages are needed. Doing so, it is possible to scale up the server without interruption on the client side. Doing so, complexity is reduced as only the server layers need to be adjusted.