Relevance Data Stores

Introduction

Goal

Configure the data stores used for different classes of data collected by the Relevance Module.

Background

The Relevance Module collects and stores different classes of data. Each has specific requirements regarding performance, reliability and querying capabilities. To meet these requirements different backend systems are supported and required.

Data Stored by the Relevance Module

The Relevance Module stores four different classes of data:

• Requests
  Raw request data. Used in the Real Time panel of the Audiences perspective as well as to derive other data such as Visits.
• Visits
  Derived from aggregated request data. Queried by the Experiments charts and Trends panel of the Audiences perspective.
• Visitors
  Data about individual visitors currently active on the site. Queried and possibly updated on every page request so its data store must be fast.
• Statistics
  Statistical data about personas and characteristics. Used to determine which persona best matches a visitor.

Each class has its own data store, configured using the Console application at the following repository nodes:

/targeting:targeting/targeting:datastores:
  /targeting:requestlog:                (requests)
  /targeting:statistics:                (statistics)
  /targeting:targetingdata:             (visitors)
  /targeting:visits:                    (visits)

Types of Data Stores

There are two different types of Data Stores used and needed for Relevance:

• SQL Database Data Store
  An SQL database is required to store Requests, Statistics and Visitors. The SQL data store can not be used to store Visits because of requirements regarding querying capabilities (see Elasticsearch below).
• Elasticsearch 5 Data Store
  Elasticsearch is used for Visits because of requirements regarding querying capabilities.