History[08][06]

Hewlett-Packard Company launched Trafodion as an open-source project in 2014, and a version was released in early 2015 before it became a part of the Apache Incubator in May 2015. In January of 2018, it was established as an Apache Top-Level Project.

The project was marked as abandoned and deprecated to the Apache Foundation "attic" in 2021.

Concurrency Control[09][10]

Apache Trafodion utilizes the HBase-Trx implementation of Multi-version Concurrency Control (MVCC). This implementation guarantees row-level ACID semantics and only requires row locks for write operations, not read operations.

Data Model[11]

Since Trafodion is built on top of HBase, it shares HBase's column family data model using key-value pairs.

Foreign Keys[11]

Foreign key constraints are supported in Trafodion.

Isolation Levels[12]

Trafodion only supports the READ COMMITTED isolation level.

Joins

Trafodion supports hash joins. It is recommended that the number of tables in a join is limited to 64, otherwise the query may not compile.

Logging[13]

Trafodion utilizes HBase's Write Ahead Log.

Query Execution[09]

Trafodion dynamically spawns a set of Executor Server Processes, which each execute a part of a query plan if intra-query parallelism is desired.

Query Interface[09]

Trafodion fully supports ANSI SQL.

Storage Architecture[14]

Trafodion is built on top of HBase, which means it shares its disk-oriented storage architecture model from HDFS.

Storage Model[14]

Trafodion uses a single column family to store all columns.

Stored Procedures[15]

Trafodion allows for stored procedures written in Java and typically compiles with SQL and JRT (Java Routines and Types). The stored procedure itself is a Java method in a JAR file.

Views

Unlike HBase, Trafodion does support views.