In Iceberg, table data consists of these parts:

• Version information: The version of the current table including the schema, partition, manifest list, and historical manifest list that corresponds to the version. 
• Manifest list: The data snapshot of a table that references a manifest file. 
• Manifest file: The aggregation of data files that provides a simple collection to merge multiple data files. 
• Data file: The original data file of the table which can be stored in Apache Parquet, Apache ORC, and Apache Avro formats.

Figure 2.   Iceberg architecture

The table metadata for Iceberg includes only the name and version information of the current table. All schemas and properties are managed by Iceberg itself. 

Iceberg supports multiple table operations. These operations include querying data from a table, inserting data into a table, updating data in a table, deleting data for a specified row, and deleting data for a specified condition. However, you can divide the process into the following two main types: modify table and query table. The following scenario uses Apache Flink as an example to introduce the two workflows of Iceberg.