EclipseLink/DesignDocs/MultiTenantFeatures

UNDER CONSTRUCTION

Purpose

EclipseLink provides a number of features that help with the challenges of developing multi tenant applications. This document will provide an overview of some of the ways EclipseLink can be used to support multi tenant applications.

In particular, this document will focus on applications where the application provider provides a core application and the tenant needs to extend the data provided in the application. (like Oracle Applications flex columns)

e.g. The application provider provides an Employee class that includes name and address. The tenant also wants to store the Employee Number. How can the application be architected to make storing that kind of extra data easy.

From a persistence-layer point of view, there are two areas where you must design to allow extra data.

1. The object model and metadata
2. The database

We will outline the options for both areas.

Object Model and Metadata

There are a number of different ways EclipseLink can support added fields in the object model.

Predefined fields

In this case, a class is designed with several generically named and typed fields. Those fields may be used by the tenant to store any data they choose. Converters could be used to control the data types coming out of the fields.

e.g. Employee contains fields called name, address, attribute1, attribute2 etc... The tenant chooses to store employee number in attribute1.

Dynamic JPA

Dynamic JPA allows the construction of a completely metadata driven application. The provider writes a dynamic JPA application that defines the set of classes and attributes that exist. The tenant calls API or provides XML that describes the additional fields on those classes.

e.g. Employee is a dynamically created entity containing name and address attributes. The tenant is allowed to construct a mapping for Employee Number by constructing an EclipseLink-orm.xml file that defines Employee with the new attribute.

Static Classes + Dynamic JPA

A static object model is provided by the provider that defines all the shared attributes. The tenant is provided with mechanism to provide extensions through Dynamic JPA. They are allowed to construct new mappings on the static classes by providing XML that describes fields that will be added to the existing object model through dynamic JPA.

e.g. Employee is a class that contains name and address attributes. The tenant is allowed to construct an eclipselink-orm.xml file that adds a virtual Employee Number field to Employee. EclipseLink's dynamic JPA functionality handles adding the field.

See https://bugs.eclipse.org/bugs/show_bug.cgi?id=331915 for the remaining work required for this configuration.

Map

Additional attributes are specified in a Map Mapping. Business logic is used to expose the fields in the Map as standard data.

e.g. Employee has attributes name and address and also a map for additional properties. To add Employee number, the tenant adds a Map Entry with key=EmployeeNumber and Value=<employees Employee Number>

Database

There are many different database architectures that can be used to support storing extra data.

Different Schema

If the tenants are all deployed on different databases or on different table spaces, the tables can quite simply be different. This can be enabled by EclipseLink's table creation feature. Additionally, it would be possible to introduce an EclipseLink feature that could alter tables in an initial schema based on different metadata.

e.g. A base tenant would have an Employee table with just a name and address field. Another tenant could have an Employee table with a name, address, and employee number field.

Static Fields

Simply define extra fields in each database table that can used to store extra data.

e.g. A table for Employee would contain name, address, attribute1, attribute2 etc...

This table architecture can be used with most of the object model strategies above.

Secondary Tables

EclipseLink supports secondary tables and these could be used to extend object models.

e.g. A base Employee table could contain name and address, a secondary table could contain a foreign key to the base table, and a set of additional field. (e.g. employee number, or attribute1 etc)

Secondary tables could be provided with preset fields in much the same way as the Static Field model above, or secondary tables could be created for each tenant that needed extra data. Dynamic metadata could be provided to change each tenants mapping to access their tables.

Map Table

Attributes are stored in a table that contains a key, an attribute name, and an attribute value. The table could optionally have a foreign key that mapped to a source table.

e.g. Employee table contains an id (and possibly some of the data). Attributes are contained in a table that contains a foreign key to the Employee table, an attribute name and an attribute value.

With this strategy, the map table could contain all the attributes, or the main table could contain some of the attributes and the map table could contain the rest.

Differentiating between tenants

EclipseLink provide 2 features that could be used to differentiate between tenants on the database

1. Support for Oracle VPD - Oracle users can use Oracle's VPD feature through EclipseLink and that feature can be used to differentiate between tenants.
2. Additional Join Criteria - EclipseLink allows additional join criteria to be specified on each descriptor that could be used for tenant differentiation

Other databases provide VPD-like features and support could be added for those as they are required.

EclipseLink Config

The following configuration options can be used to configure the different stratgies

Single Persistence Unit

A single persistence unit could be provided that maps all the attributes. This strategy is useful in the case where there is no dynamic component to the application, all the db fields and all the attributes remain the same.

Deployment per tenant

A full persistence unit could be configured per tenant. That persistence unit could contain different persistence.xml, mapping files, classes and resources. This could be used with any of the above strategies.

Persistence xml per tenant

In this case, each tenant has a persistence.xml file which creates them a separate persistence unit. The classes used by each tenant remain the same. EclipseLink supports dynamic selection of persistence.xml so as long as the persistence xml for each tenant could be provided, this could all be configured in one deployment.

eclipselink-orm.xml per tenant

It is possible to add a persistence unit property that allows orm.xml files to be provided by on a per-tenant basis. In this case, everything about the persistence unit would be the same except a single eclipselink-orm.xml file which defines the mappings that are particular to a single tenant.

Customizer per tenant

A customizer could be used to configure tenant data. Each tenant would have a customizer that cusomized their persistence unit.