High availability limitations

Alongside benefits, keep in mind some limitations of a highly available AVEVA™ PI System™.

• All servers and interfaces must be in a single Active Directory domain

  We designed the PI System to support high availability in environments with all servers and interfaces in a single domain—a domain configured with a domain controller and a reliable DNS (domain name system) resolution. You must use special configuration procedures if:

• You have components not installed in a homogeneous security environment, such as components installed in different, non-trusted domains, or components installed in a work group.

• You do not have access to Active Directory (AD) and must configure authentication through local Windows security.

• Data Archive servers distributed geographically

  For enterprise-wide Data Archive servers that are distributed geographically, a PI to PI Interface instead of Data Archive collectives is better suited because it is likely that different sites use different security models. The PI to PI interface transfers data from one Data Archive server to another Data Archive server via TCP/IP.

• Collective Manager requires Windows file copy access

  You can easily create Data Archive collectives and manage servers those collectives with PI Collective Manager. However, PI Collective Manager requires Windows file copy access between servers. This requires properly opened TCP ports. Without this access, you must manually create collectives and initialize secondary servers.

• Not all data is replicated

  Some data is only written to the primary Data Archive server, so if the primary goes down, you need to recover data from the primary, and not simply promote a secondary.

The PI System uses the buffer mechanism to replicate data from interfaces to the servers in a Data Archive collective. Therefore, data not sent to Data Archive server through the buffering system is not replicated.

• No replication of batch records

  The PI Batch Database identifies objects by a unique ID. Each Data Archive server randomly assigns a unique ID to a batch object. Therefore, each collective member will generate a unique ID for batch objects with the same data and configuration. Though you think it is the same object, the software will interpret them as different objects.

• Performance Equation Scheduler limitations

  Performance Equation Scheduler is not aware of high availability features: it can interact with only one Data Archive server. Because all servers in a Data Archive collective will have the same input tags to a Performance Equation, results will be the same in most cases, regardless of which server the scheduler connects to. However, buffering and network connection issues can introduce variation.

To avoid any variation, use applications that are aware of Data Archive collectives, like the calculation functions of PI AF.