Data Stores
What is a Data Store?
A data store is a repository for storing, managing, and distributing data sets.
The term includes all types of data that are produced, used, and stored by an organization.
What are the different types of data stores?
The following are common types of data stores.
Relational database (RDB): The most durable and reliable type of data store, a relational database is the industry standard for reliable storage. An RDB organizes data into tables, each having a schema that defines the columns for a table; each row, representing a record of information, must conform to the schema by having a value for each column. In short, information is assigned a schema value, thus a relationship between value and information is established and maintained.
Non-relational database (NoSQL): A database that maintains durability, resilience, persistence, replicability, distributability, and performance while not enforcing (or loosely enforcing) schemas. NoSQLs are subcategorized into two primary categories: document stores and wide column stores.
- Document store: A key-value store where the key is never used. The value becomes a blob of semi-structured data, and the data store is a big array of blobs. The language of the document store allows a user to sort or filter data based on the content found in the document blobs.
- Wide column store (WCS): Essentially a hybrid document store and RDB. Though a WCS uses tables, rows, and columns, the names and formats of the columns form rows in a single table.
Key-value store: A production-scale hashmap (a hashmap being a map from keys to values). This data store type has no relational or non-relational elements, only keys and values. This type is good for storing simple objects temporarily.
Full-text search engine (FTSE): Technically FTSEs are NoSQL data stores. While search engines are good at searching and filtering by exact text matches and numeric values (DBs can achieve similar results), FTSEs are ideal for looking for specific substrings or words within longer text fields.
Message queue: While originally intended for data transfer, message queues perform as reliably as the earlier data store types. A message queue performs as a pseudo-key-value store, but it is best used when you need to temporarily store, queue, or ship data.
What does a data store include?
A data store may include data from end user database applications, random data property of an organization or information system, files, or documents. It may be structured, unstructured, or in a variety of electronic formats.
Classification of a data store depends on the organization. A data store may be classified as a centralized, operational, or application-specific data store, and may be designed and implemented by using purpose-built software or through typical database application.
HPE and data stores
A single data store is a logical idea. The best way to execute a single data store is using a logical deployment. HPE Ezmeral is a logical idea: an open-platform, high-performance object store, multi-modal database, hybrid multi-tenancy, and a single global namespace.
Into a single data store, HPE Ezmeral integrates files, NoSQL databases, objects, and multiple types of streaming data from cloud-native architectures and existing big data. This allows users to accelerate time to insights and secure data sharing across modern and traditional data analytics apps and tools.
HPE Ezmeral Data fabric’s edge-to-cloud topologies are created and accessed via a single global namespace simplifying data access from any application, regardless of the interface or where the data resides. HPE Data Lakehouse is specifically designed not only to perform large amounts of structured data analytics, but also to combine scalability and flexibility benefits of HPE Data Lake with data management and structures of the Data Warehouse.