Issue link: https://resources.awscloud.com/i/1496542
©2020 IDC #US47078920 3 Multimodel and Broadly Featured DBMSs Today, there are many DBMSs that have developed multimodel capabilities; that is, they can support more than one data management model. For instance, an RDBMS may also support document data, key-value data, and graph data, in addition to relational data. For the most part, these multimodel DBMSs are generally very good at the main job for which they were designed, but their capabilities in the other models of operation lack the performance, scalability, and ease of development and use offered by database systems that are purpose built to support those specific models. This does not mean that such DBMSs have no value in these areas. Often, an organization may want to keep its business data in relational tables but maintain connections with JSON document kept in the same database for the sake of data integrity, for instance. This is a perfectly reasonable use of such systems, where the primary operational model is relational. Workload-Targeted DBMSs In an environment in which data can be shared among databases in a carefully controlled and well- governed way, it is often preferable to use a DBMS that is targeted at a particular workload, such as relational data management, or document management, or graph analysis, and excels at supporting that workload. Relational databases can organize and optimize tables in row or columnar format for the most efficient transaction processing and fastest query processing. Document databases can optimize document structures and storage to provide the most efficient retrieval and update combined with high- speed search. Key-value stores can optimize for very high-speed randomized storage and retrieval. Graph databases can provide an environment optimized for either broad and shallow or narrow and deep graph traversal and for the management of either property or Resource Description Framework (RDF) graphs. Choosing the Right DBMS for the Job For applications that are more ideally targeted at a particular type of database workload, however, it generally makes more sense to choose a DBMS that is purpose built for that workload in order to get the best result in terms of high availability, low latency, extreme performance, and behavior that is well attuned to the particular type of data in question. Table 1 illustrates the various database models, how they are used, and examples of common workloads they serve.