Relational Systems (1970)

Edgar Codd worked at IBM, in one of those peripheral offices that was mainly dedicated to the development of hard drives. I was unhappy with the CODASYL navigation model, mainly the lack of search operation. In 1970 he wrote some articles in which he outlined a new approach that culminated in the document "A Relational Model of Data for Large Shared Data Banks".[3]​.

In this article he discovered a new system for storing and working with large databases. Instead of storing records of arbitrary type in a chained list as in CODASYL, Codd's idea was to use a "table" of records of fixed size. A chained list has very little efficiency when storing sparse data where some of the data in a record may be left blank. The relational model solves this by dividing the data into a series of normalized tables—or relationships—in which optional elements have been extracted from the main table so that they take up space only if they need it. In this relational model related records are linked with a "key".

A common use of databases can maintain an address book of users, their name, login information, address and telephone number. In the navigation solution all that data would be located in a single record, and unused features would simply not be in the database. In the relational solution, the data would be normalized in a user table, a telephone table and an address table, in which records would be added if we had to incorporate telephone and address.

Reconciling all information is the key to this system. In the relational model, a part of the information is used as a key, uniquely identifying a specific record. When information is collected about a user, information in the optional tables will be accessed by searching using that key. For example, if the username is unique, the address and phone number of that user will be saved with the username as a password. The compilation of this information in a single record is something that traditional languages ​​are not designed for.

Just as the navigation approach requires programs that loop to collect records, the relational approach will also require them. Codd's solution to the necessary loops is based on a set-oriented language, a suggestion that would later crystallize into the ubiquitous SQL. He proposed the use of a branch of algebra called tuple calculus, and demonstrated that with it all typical operations could be performed on a database, in addition to extracting data sets in a simple way.

Codd's article fell into the hands of two people in Berkeley, Eugene Wong and Michael Stonebraker. They started a project called INGRES with funds allocated to a geographic database project programmed by the students. Starting in 1973, INGRES produced its first test versions that were ready for general use in 1979. INGRES was very similar to IBM's System R in several ways, including a language for data access, known as QUEL. Over time, INGRES adopted the SQL standard.

IBM carried out a test implementation of the relational model - PRTV - and a production one - Business System 12 - both of which were discontinued. Honeywell wrote MRDS for Multics, and two new implementations also appear: Alphora Dataphor") and Rel. Most other implementations of so-called relational DBMSs are actually SQL DBMSs.

In the 1970s, the University of Michigan began the development of the MICRO Information Management System based on D. L. Childs' theoretical data model. Micro was used to manage large amounts of data in the Department of Labor of the US government. It ran on mainframe using Michigan Terminal System. It was in production until 1998.

SQL systems (late 1970s)

IBM began work in the early 1970s on a prototype loosely based on Codd's concepts, calling it System R. The first version was ready in 1974 or 1975, and work on multi-table systems began, in which data could be disaggregated so that all the information in a record (some of which is optional) did not have to be stored in a single large chunk. Subsequent multi-user versions were tested by users in 1978 and 1979, by which time an SQL language had been standardized. Codd's ideas were revealed to be operational and superior to those of CODASYL, launching IBM into the development of a true production version of System R, known as SQL/DS, and later as Database 2 (DB2).

Many of the INGRES technicians were sure of the commercial success of the system, and formed their own companies to commercialize the development but with an SQL interface. Sybase, Informix, NonStop SQL and INGRES itself were sold as derivatives of the original INGRES in the 1980s. Even Microsoft's SQL Server is based on Sybase, and therefore on INGRES. Only Larry Ellison—the founder of Oracle—started a new path based on IBM's article on System R, and surpassed IBM by releasing its first version in 1978.

Stonebraker applied the lessons of INGRES to the development of a new database—Postgres—now known as PostgreSQL. PostgreSQL is used for many critical applications (.org and .info domain registries use it for their primary storage, as well as large companies and financial institutions).

In Sweden, Codd's paper generated the Mimer SQL database[4]​ at Uppsala University. In 1984 this project was consolidated into an independent company. In the early 1980s, Mimer introduced transaction management to provide robustness to applications, an idea that was picked up in many other DBMSs.

Object-oriented systems (1980)

During the 1980s, the rise of object-oriented programming influenced the way information in databases was handled. Programmers and designers began to treat data in databases as objects. This means that if a person's data is in the database, the person's attributes such as address, phone number, and age are considered to belong to the person, not foreign data. This allows relationships to be established between objects and attributes, rather than between individual fields.

Another great focus of attention during the decade was the increase in speed and reliability of access. In 1989, two University of Wisconsin professors published a paper at an ACM conference outlining their methods for improving database performance. The idea was to replicate the important—and most requested—information in a small temporary database with links to the main database. This meant that the small database could be searched much faster than the large one. Its improved performance led to the introduction of indexing, incorporated into all DBMSs.

NoSQL Systems (2000)

The century brought a new trend in databases: NoSQL. This trend introduced a non-relational line significantly different from the classic ones. They generally do not require fixed schemas, avoid join operations by storing denormalized data, and are designed to scale horizontally. Most of them can be classified as key-value stores or document-oriented databases.

Recently there has been a great demand for distributed databases with partition tolerance, but according to the CAP theorem it is not possible to achieve a distributed system that simultaneously provides consistency, availability and partition tolerance. A distributed system can satisfy only two of the three constraints at a time. For this reason many NoSQL databases use so-called eventual consistency to provide availability and partition tolerance, with a maximum level of data consistency.

Among the most popular applications we find MongoDB, MemcacheDB, Redis, CouchDB, Hazelcast"), Apache Cassandra and HBase, all of which are open source.

XML Systems (2010)

XML databases form a subset of NoSQL databases. They all use the XML storage format, which is open, human- and machine-readable, and widely used for interoperability.

In this category we find: BaseX, eXist, MarkLogic Server, MonetDB/XQuery, Sedna.

Navigation systems (1960)

As computers gained speed and capacity, general-purpose database systems appeared; By the mid-1960s there were some systems in use. Interest in obtaining a standard arose and Charles Bachman—author of one of the first products, the Integrated Data Store (IDS)—founded the Database Task Group within CODASYL, the group responsible for the creation and standardization of COBOL. In 1971 they published their standard, which became known as the "CODASYL approach", and soon some products based on this line appeared.

CODASYL's strategy was based on manual navigation through a network-linked data set. When the database was started, the program returned a link to the first record in the database, which in turn contained pointers to other data. To find a specific record, the programmer had to follow pointers until reaching the desired record.

To answer simple questions like “find all people in Japan,” the program had to go through all the data to choose the correct records. There were no concepts of “search” or “find”, something that would be unacceptable today, but in the days when data was stored on tape it was not feasible to put them into practice.

Solutions were found to many of these problems. The manufacturer Prime created a CODASYL-compliant DBMS based on binary trees that shortcutted navigation from record to record by providing alternative access paths. It also provided very clear query language. In fact there is no reason not to be able to apply normalization concepts to CODASYL databases, but ultimately CODASYL was very complex and required a lot of effort and practice to produce a useful application.

IBM also had its own DBMS in 1968, known as IMS "IMS (IBM)"). It was software developed for the Apollo program on System/360. IMS had similar concepts to CODASYL, but used a strict hierarchy of data ordering, compared to the network structure of CODASYL. Both concepts were later included in the concept of navigation databases due to the way the data was accessed. In fact, Bachman received the Turing Award in 1973 for his presentation The programmer as a navigator.[2]​.

Relational Systems (1970)

Edgar Codd worked at IBM, in one of those peripheral offices that was mainly dedicated to the development of hard drives. I was unhappy with the CODASYL navigation model, mainly the lack of search operation. In 1970 he wrote some articles in which he outlined a new approach that culminated in the document "A Relational Model of Data for Large Shared Data Banks".[3]​.

In this article he discovered a new system for storing and working with large databases. Instead of storing records of arbitrary type in a chained list as in CODASYL, Codd's idea was to use a "table" of records of fixed size. A chained list has very little efficiency when storing sparse data where some of the data in a record may be left blank. The relational model solves this by dividing the data into a series of normalized tables—or relationships—in which optional elements have been extracted from the main table so that they take up space only if they need it. In this relational model related records are linked with a "key".

A common use of databases can maintain an address book of users, their name, login information, address and telephone number. In the navigation solution all that data would be located in a single record, and unused features would simply not be in the database. In the relational solution, the data would be normalized in a user table, a telephone table and an address table, in which records would be added if we had to incorporate telephone and address.

Reconciling all information is the key to this system. In the relational model, a part of the information is used as a key, uniquely identifying a specific record. When information is collected about a user, information in the optional tables will be accessed by searching using that key. For example, if the username is unique, the address and phone number of that user will be saved with the username as a password. The compilation of this information in a single record is something that traditional languages ​​are not designed for.

Just as the navigation approach requires programs that loop to collect records, the relational approach will also require them. Codd's solution to the necessary loops is based on a set-oriented language, a suggestion that would later crystallize into the ubiquitous SQL. He proposed the use of a branch of algebra called tuple calculus, and demonstrated that with it all typical operations could be performed on a database, in addition to extracting data sets in a simple way.

Codd's article fell into the hands of two people in Berkeley, Eugene Wong and Michael Stonebraker. They started a project called INGRES with funds allocated to a geographic database project programmed by the students. Starting in 1973, INGRES produced its first test versions that were ready for general use in 1979. INGRES was very similar to IBM's System R in several ways, including a language for data access, known as QUEL. Over time, INGRES adopted the SQL standard.

IBM carried out a test implementation of the relational model - PRTV - and a production one - Business System 12 - both of which were discontinued. Honeywell wrote MRDS for Multics, and two new implementations also appear: Alphora Dataphor") and Rel. Most other implementations of so-called relational DBMSs are actually SQL DBMSs.

In the 1970s, the University of Michigan began the development of the MICRO Information Management System based on D. L. Childs' theoretical data model. Micro was used to manage large amounts of data in the Department of Labor of the US government. It ran on mainframe using Michigan Terminal System. It was in production until 1998.

SQL systems (late 1970s)

IBM began work in the early 1970s on a prototype loosely based on Codd's concepts, calling it System R. The first version was ready in 1974 or 1975, and work on multi-table systems began, in which data could be disaggregated so that all the information in a record (some of which is optional) did not have to be stored in a single large chunk. Subsequent multi-user versions were tested by users in 1978 and 1979, by which time an SQL language had been standardized. Codd's ideas were revealed to be operational and superior to those of CODASYL, launching IBM into the development of a true production version of System R, known as SQL/DS, and later as Database 2 (DB2).

Many of the INGRES technicians were sure of the commercial success of the system, and formed their own companies to commercialize the development but with an SQL interface. Sybase, Informix, NonStop SQL and INGRES itself were sold as derivatives of the original INGRES in the 1980s. Even Microsoft's SQL Server is based on Sybase, and therefore on INGRES. Only Larry Ellison—the founder of Oracle—started a new path based on IBM's article on System R, and surpassed IBM by releasing its first version in 1978.

Stonebraker applied the lessons of INGRES to the development of a new database—Postgres—now known as PostgreSQL. PostgreSQL is used for many critical applications (.org and .info domain registries use it for their primary storage, as well as large companies and financial institutions).

In Sweden, Codd's paper generated the Mimer SQL database[4]​ at Uppsala University. In 1984 this project was consolidated into an independent company. In the early 1980s, Mimer introduced transaction management to provide robustness to applications, an idea that was picked up in many other DBMSs.

Object-oriented systems (1980)

During the 1980s, the rise of object-oriented programming influenced the way information in databases was handled. Programmers and designers began to treat data in databases as objects. This means that if a person's data is in the database, the person's attributes such as address, phone number, and age are considered to belong to the person, not foreign data. This allows relationships to be established between objects and attributes, rather than between individual fields.

Another great focus of attention during the decade was the increase in speed and reliability of access. In 1989, two University of Wisconsin professors published a paper at an ACM conference outlining their methods for improving database performance. The idea was to replicate the important—and most requested—information in a small temporary database with links to the main database. This meant that the small database could be searched much faster than the large one. Its improved performance led to the introduction of indexing, incorporated into all DBMSs.

NoSQL Systems (2000)

The century brought a new trend in databases: NoSQL. This trend introduced a non-relational line significantly different from the classic ones. They generally do not require fixed schemas, avoid join operations by storing denormalized data, and are designed to scale horizontally. Most of them can be classified as key-value stores or document-oriented databases.

Recently there has been a great demand for distributed databases with partition tolerance, but according to the CAP theorem it is not possible to achieve a distributed system that simultaneously provides consistency, availability and partition tolerance. A distributed system can satisfy only two of the three constraints at a time. For this reason many NoSQL databases use so-called eventual consistency to provide availability and partition tolerance, with a maximum level of data consistency.

Among the most popular applications we find MongoDB, MemcacheDB, Redis, CouchDB, Hazelcast"), Apache Cassandra and HBase, all of which are open source.

XML Systems (2010)

XML databases form a subset of NoSQL databases. They all use the XML storage format, which is open, human- and machine-readable, and widely used for interoperability.

In this category we find: BaseX, eXist, MarkLogic Server, MonetDB/XQuery, Sedna.