How do Document Stores work?

Databases are necessary for or­gan­ising in­form­a­tion in a practical manner. However, there are various ways for databases to be struc­tured. In elec­tron­ic data pro­cessing, re­la­tion­al databases are par­tic­u­larly common and wide­spread. Besides these, there are document-based databases. These are based on a simple table structure with documents for storing in­form­a­tion. How do these databases work and what are their ad­vant­ages?

Document-oriented databases, also known as document stores, are used to manage semi-struc­tured data. This data does not adhere to a fixed structure, instead it forms its own structure. The in­form­a­tion can be ordered using markers within the semi-struc­tured data. Due to the lack of a defined structure, this data is not suitable for re­la­tion­al databases since its in­form­a­tion cannot be arranged in tables.

A document database creates a simple pair: A key is assigned to a specific document. The actual in­form­a­tion is then located within this document, which may be formatted as an XML, JSON or YAML file. Since the document does not require a specific schema, different types of documents can also be in­teg­rated together in a document store. Changes to the documents do not have to be com­mu­nic­ated to the database.

In theory, data in all sorts of formats, even without a con­sist­ent schema can be stored in a document-based database. In practice, however, a file format is typically used for the documents and the in­form­a­tion is ordered in a certain structure. This makes it easier to work with the in­form­a­tion and database. By using data struc­tures, database search queries can be processed more ef­fect­ively for example. You can generally perform the same actions in a document-based database as with a re­la­tion­al system: in­form­a­tion can be added, changed, deleted, and queried.

To allow these actions to take place, each document is given a unique ID. How this iden­ti­fi­er is con­sti­tuted is not par­tic­u­larly important. Both a simple number series, or the complete pathway can be used to address the document. When searching for in­form­a­tion, the documents them­selves are checked. In other words, the data is pulled directly from the documents rather than from the columns within the database.

In con­ven­tion­al re­la­tion­al databases, a field has to exist for each piece of in­form­a­tion—and in every entry. If the in­form­a­tion is not available, the cell is kept empty, but it must still exist. Document-oriented databases are much more flexible: the structure of in­di­vidu­al documents does not have to be con­sist­ent. Even large volumes of un­struc­tured data can be ac­com­mod­ated in the database.

Plus, it’s easier to integrate new in­form­a­tion. While in the case of a re­la­tion­al database a new in­form­a­tion criterion must be added to all datasets, the new in­form­a­tion only needs to be included in just a few datasets in a document store. The ad­di­tion­al content can be added to further documents, but it’s not required.

Moreover, with document stores the in­form­a­tion is not dis­trib­uted over multiple linked tables. Everything is contained in a single location, and this can result in better per­form­ance. However, this speed advantage is only realised in document databases so long as you don’t attempt to use re­la­tion­al elements: ref­er­ences don’t really suit the concept of document stores. If you do try to interlink the documents, the system will become highly complex and cum­ber­some. So, a re­la­tion­al database system is more advisable for highly networked data volumes.

Es­pe­cially for the de­vel­op­ment of web apps, databases for documents are hugely important. Due to the increased need resulting from web de­vel­op­ment, numerous database man­age­ment systems (DBMSs) have meanwhile been released on the market. The most well-known examples are outlined below:

• BaseX: This open-source project uses Java and XML. BaseX is supplied with a graphical user interface.
• CouchDB: The Apache Software Found­a­tion released the open-source software CouchDB. The database man­age­ment system is written in Erlang, uses JavaS­cript, and is utilized in Ubuntu and Facebook ap­plic­a­tions among others.
• Elast­ic­search: This search engine works based on a document-oriented database. JSON documents are used to this end.
• eXist: The open-source DBMS runs on a Java virtual machine and can therefore be used re­gard­less of the operating system. XML documents are primarily used.
• MongoDB: MongoDB is by far the most wide­spread NoSQL database. The software is written in C++ and uses JSON-like documents.
• SimpleDB: With SimpleDB (written in Erlang), Amazon developed its own DBMS for the company’s Cloud services. The provider charges a fee for use.