Multi-cluster architectures

A multi-cluster architecture lets you scale your system as your needs evolve. A multi-cluster architecture is intended for use in large-scale, distributed SCADA systems in which multiple clusters of servers and clients work together under a single project.

When to use a multi-cluster architecture

A multi-cluster architecture should be considered when:

  • The project exceeds 300,000 tags, or growth is anticipated beyond this.

NOTE: The number of tags required before a multi-cluster architecture is needed depends on the makeup of the project, the number of devices, and the project configuration. A multi-cluster architecture may be required before the number of tags exceeds 300,000.

  • The project exceeds total CPU and/or memory usage > 60%.
  • The Trend Server's 32-bit Citect process reaches > 3 GB of process memory usage.
  • The system has different sets of runtime components (alarms, trend, report servers) within a single project, or spans multiple sites/geographical areas that can be monitored and controlled through a single client.

Advantages and disadvantages of multi-cluster architecture

Advantages:

  • Scalability: A multi-cluster architecture improves the handling of more tags, devices, alarms, trends, etc.
  • Performance: A multi-cluster architecture improves the distribution of load for alarms, trends, and I/O across multiple servers.
  • Fault isolation: A multi-cluster architecture is designed so that issues in one cluster typically do not affect others.
  • Redundancy: Each cluster may be configured to have its own redundant servers.
  • Flexibility: Operational clusters can be set up to run independently.

Disadvantages:

  • Complexity: A multi-cluster architecture requires careful configuration of networking and synchronization.
  • Higher cost: A multi-cluster architecture requires more hardware, licenses, and maintenance.
  • Troubleshooting difficulty: Diagnosing issues across clusters is more involved.

Performance impact of moving to multi-clusters

Due to the 32-bit Citect processes, each process has a memory limit of up to 3.5 GB. Moving from a single cluster to two clusters distributes load across multiple servers. Each cluster gets its own Alarm and Trend Server, which can help with reducing memory and CPU stress on the single server and improving the trend and alarm retrieval.

NOTE: Multi-cluster configuration will not resolve performance-related issues due to improper loading of I/O servers and pre-existing issues within the system related to misconfiguration.

Available architectures

You can use the following architectures:

  • Multi-site project: Multiple individual PO projects (normally configured for individual sites) are included together within a master project. Each of the individual PO projects can be configured to have one or more clusters.
  • Multi-cluster project: A project consisting of two or more clusters.

Use the information in the tables below to find the content you are looking for:

Topic Description
Server architecture Example of a multi-cluster architecture.
SCADA project structure SCADA project structure for a multi-clustered system.
Project development structure Example project development structure for a multi-clustered system.
Configuration guidelines Guidelines for configuring multi-clustered systems.
Multi-cluster architecture example projects

Working examples demonstrating how to set up and run multi-site and multi-cluster projects.