How does FeatureBase differ to a traditional database?

FeatureBase is a database that uses a bitmap-based data format rather than pure bitmap indexes.

This high-level overview provides context by explaining:

• data cardinality
• data normalization in traditional databases

What is cardinality?

The relationships between data is called Cardinality and can be conceptualized as follows:

High cardinality data

High cardinality data has a high number of unique relationships which can be represented in a two dimensional table:

Low cardinality data

Data described as low cardinality have multiple relationships has a one-to-many or many-to-many relationship. For example:

Database normalization in relation to data cardinality

Database normalization has a set of normal forms which provide guidance on how data is represented.

The first normal form provides guidance on:

• arranging data into two dimensions
• the use of relation names, attributes and keys to reference rows

This means:

• high and low cardinality data is saved to separate tables to remove duplication
• the relationships (one-to-many and many-to-many) are maintained through the use of keys that reference specific rows in different tables.

For example, the low cardinality table above can be normalized as follows:

The SubjectID can then be linked with a key to the Students table StudentID key.

Benefits and costs of data normalization

Data normalization is not a perfect solution to data cardinality:

DBAs responsible for normalized systems use different methods to overcome the issues and should the benefits outweigh the costs, may denormalize data.

How does FeatureBase handle data cardinality?

FeatureBase does not use Database normalization. Instead, the system inserts data into a two-dimensional bitmap index which is:

• designed to overcome issues with low cardinality data
• optimized to reduce storage overheads and query execution time

Learn how data is encoded in bitmap indexes

How should I structure data to be imported to FeatureBase?

Data modeling in FeatureBase involves:

• identifying the issues you’re experiencing with your source data
• identifying the data to import and the unique key for each row
• mapping data types, including those that handle high-cardinality data
• choosing a method to import your data
• creating the destination
• running the import
• testing the outcome
• fixing issues that may have occurred

Learn how to perform Data Modeling for the FeatureBase Bitmap database

Examples with data

• Data import examples

Further information

• Cardinality(data modeling) on Wikipedia
• Database Normalization on Wikipedia
• Database denormalization on Wikipedia

---

Table of contents

• Bitmap index overview
• Glossary
• Data modeling overview
• Examples

---