Unisys DMSII
| Industry | Computer hardware and software |
|---|---|
| Founded | 1953 |
| Fate | Merged with Sperry Corporation to form Unisys |
| Successor | Unisys |
| Products | Unisys Data Management System II (DMSII) |
The Unisys Data Management System II (DMSII) is a proprietary database management system initially created by the Burroughs Corporation in 1972. It was designed to run on Burroughs (which later became Unisys) Small (B1000), Medium (4000, V Series), and Large System (5000, 6000, 7000) hardware families. Subsequent support extended to A Series, ClearPath, and Libra platforms. In more recent Unisys ClearPath editions, it has been rebranded as the Enterprise Database Server for ClearPath MCP. The original implementation of DMSII was based on a network database architecture.
DMSII was positioned as a competitor to IBM's IMS and CICS offerings before relational databases became dominant.
History
DMSII offered an ISAM (indexed sequential access method) approach for data retrieval, transaction isolation, and recovery processes. Its schema was defined using a proprietary language known as Data and Structure Definition Language (DASDL). The DASDL source code compiled into a metadata descriptor file, which was then utilized by a utility to generate the necessary database components.
Compilers for programming languages such as COBOL and ALGOL used this metadata file to incorporate database constructs directly into the object code via language extensions, allowing tight integration between applications and database structures.
Originally, DMSII maintained a data dictionary referenced by application source code stored within CANDE. The DASDL compiler either updated or created this dictionary and automatically produced programs to facilitate data migration during structural modifications. Developers only needed to recompile their source code against the revised description file to access newly defined data elements.
Data was organized into "sets," which contained records that could be accessed randomly or in sequence. These sets supported both hierarchical and relational data models and could be either disjoint or embedded. Indexes were realized as index sets—either complete or subsets—with subsets classified as automatic (rule-driven) or manual (program-controlled).
Index sets included ISAM types based on field keys as well as bit-vectored sets that represented membership through Boolean logic. This approach enhanced data retrieval speed but introduced additional overhead when inserting records associated with numerous subsets.
Manual subsets enabled the construction of network-style databases but demanded meticulous programmatic control to preserve referential integrity. Unlike relational databases where indexes can be dynamically created during runtime, DMSII required indexes to be predefined or established during database reorganizations.
In 1988, Burroughs launched Infoexec, an object-oriented database layer built atop DMSII and implemented in Pascal. Developed by Mike Hammer and Dennis McLeod at MIT, Infoexec sought to provide runtime selection capabilities similar to those found in relational systems but ultimately suffered from suboptimal performance.