System Architecture

Standard or Stand-Alone Historian Architecture:

In this type of system, there is a single Historian server. It offers the following unique capabilities and benefits for a sustainable competitive advantage:
  • Built-in data collection
  • Good read/write performance speed
  • Enhanced data security
The following image shows the architecture of a stand-alone Historian system (single server):

Horizontally Scalable Historian system:

In addition to the capabilities of a stand-alone Historian system, a horizontally scalable one offers data redundancy and high availability. You can have mirroring of stored data on multiple nodes to provide high levels of data reliability. Data mirroring also involves the simultaneous action of every insert, update, and delete operations that occur on any node. You can spread the data collection, server, administration, and client data retrieval functions across various nodes.

In a typical data mirroring scenario, one server acts as a primary server to which the clients connect. To create a mirror, you must add mirror nodes and establish a data mirroring session relationship between the server instances. All communication goes through Client Manager, and each Client Manager knows about the others.

When a client (either a writing collector or reading client) connects to the Client Manager, it gathers information about each Client Manager, along with all archive, tag, and collector configuration information, from the Configuration Manager, and stores this information locally in its Windows Registry.

A relationship is then established between each remote client and a single Client Manager, which directs read and write requests across the other mirrors. If that relationship is broken, it will establish a new relationship with the next available Client Manager, which assumes the same responsibilities. This bond is maintained until that Client Manager is unavailable, and then the process of establishing a relationship with another Client Manager is repeated.

When more than one node is running, the Client Manager uses a "round robin" method between the good nodes to balance read loads. Each read request is handled by a node as a complete request.

Writes are sent independently but nearly simultaneously to any available data archiver so that the same tag shares a common GUID, name, timestamp, value, and quality as passed to it by the collector.

The following diagram shows the architecture of a horizontally scalable Historian system.