Issue link: https://resources.awscloud.com/i/1496542
December 2020, IDC #US47078920 White Paper Choosing the Right Tool for the Job: Purpose-Built Databases from AWS Sponsored by: Amazon Web Service Carl W. Olofson December 2020 IN THIS WHITE PAPER The most commonly used database management systems (DBMSs) are designed to support relational databases. Often, additional functionality has been folded into them to provide some support for other operational models. Their support for these operational models usually proves inadequate for work that is centered on non-relational models. For such operational models, including document database, graph analysis, or other specialized models of operation, a better choice is to use a DBMS that is specifically designed to address each such model of operation. This white paper examines the problem of picking the right DBMS for the operational model at hand. It considers how generalized multimodel DBMSs often are jacks of all trades but masters of none, while a specific, purpose-built DBMS is the better choice for a specific data management problem. It then looks at the range of purpose-built DBMSs offered by AWS, examining how each is aimed at a specific data management capability to deliver maximum value for that model. SITUATION OVERVIEW Database management systems have been in use since the early 1960s. They overcame the limitations of fixed batch jobs and flat files by providing data governed according to an application- neutral structure, called a "schema," so that the problem of file management was greatly reduced and application programs could be run in any order. The problem with these DBMSs was that they were difficult to query unless one had exquisitely precise knowledge of their structure. Dr. Edgar F. Codd, an IBM Fellow, overcame this problem with relational data management, which calls for the organization of data according to mathematical set theory. Greatly simplified, this model calls for unique keys and some number of associated attributes forming a relation. Every list of a key value and its associated attributes is called a tuple. For common simplicity, we refer to a relation as a table, a tuple as a row, and an attribute (as well as the key) as a column. A system of organization that ensures consistency without either ambiguity or redundancy is normally used, which is called "normalization." Table rows may be related to rows in other tables through a foreign key value that matches the primary key value of the other table. A standard query language emerged to support it, called Structured Query Language (SQL). As new DBMSs were developed based wholly on relational data management, SQL was expanded to include data manipulation as well as query.