Skip to content

OrderFlow Technical Architecture

OrderFlow Ltd.

Document Version: 4.2.4

Document Built: 2024-02-16

This document and its content is copyright of OrderFlow Ltd. All rights reserved.
You may not, except with our express written permission, distribute, publish or commercially exploit the content.
Any reproduction of part or all of the contents in any form is prohibited.

Data View

Data View

As detailed in the Data Layer section, OrderFlow consists of one or more relational databases, running in a MariaDB server.

To give an idea of the size of OrderFlow, it's main database has (at the last count) 167 tables, plus another 76 views.

Tables for data entities are created with auto-incrementing identifiers, to allow the application to unambiguously identify entity records. Some of these entities also have external reference fields, which hold a reference to the entity that may be used outside of OrderFlow. For example, an order's external reference would be known by (indeed supplied by) a shopping cart system.

Relationships are maintained by foreign key associations (for simple 'many-to-one' relationships), or join tables (for the more complex 'many-to-many' relationships). This ensures that the application's data cannot lose its integrity.

Some entity records are never deleted from the system, since they hold some referential integrity for other entity records. For example, a location is not deleted if there have been stock changes that refer to that location. Instead, if the user wishes to delete such an entity, its deleted flag is set, and it is never subsequently used by the application.

Indexes are applied to certain columns in specific tables. This speeds up data access for the specific queries that the system queries performs.


Changes to entity tables are recorded with varying levels of traceability. Some changes just record the time at which the change was made, whereas others additionally record which user made that change. This can be useful when determining when and who made changes to both transactional and configuration entities on the system.


To avoid a near-concurrent update of an entity record overwriting an earlier update (by a different user), entity records can be versioned. This allows the application to adopt an optimistic locking strategy with regard to data updates - an entity record's version is checked before each update and incremented if it differs from the expected value.