Performance Guidelines

Describes how to get the best performance from your App Connect for Manufacturing system.

Apply adapter pattern

It is a common best practice in integration to logically separate integration flows in three parts: input, common transformation, and output. The central processing can be applied on a common data model individually to each of the submodel parts for re-usability. And the transformation from or to the integrated system-specific models is logically separated.

It is advisable to use the ACMfg nodes in separate flows, application runtime, or containers. For example, a flow gets the subscribed items through an ACMfg Input node and puts them into an MQTT topic or MQ queue, which is the starting point of another flow that does the actual route, transform and enrichment. It enables the transaction ability, isolation, and re-usability of the integration functionality.

Number of connected OPC UA endpoints

Be cautious with the number of OPC UA servers that are connected in your ACMfg deployment for the following reasons:

  • The configuration effort increases in an ACMfg environment by the number of OPC UA data sources. For example, synchronizing the OPC UA URLs.
  • The management effort increases due to the follow-on errors at ACMfg that results from OPC UA server disconnection. It is difficult to distinguish between a real error or a normal maintenance scenario at the OPC UA server.
  • The load test effort becomes stringent to simulate a large number of OPC UA test clients.

    Having several dozens of connected OPC UA would intensely increase the manageability of the whole deployment before the use of the OPC UA publish/subscribe specification. You can use the MQTT connectivity that the OPC UA servers are providing in parallel to the OPC UA server capability.

    For example, Siemens S7-1500, Kepware, telit deviceWISE, or Softing.

Optimum load
Stable performance is observed with an optimum load of 30 datasources that each has 50 monitored items with minimal processing in the flow.
Note:
  • This configuration is recommended as the optimum load to retain a stable performance while you are using ACMfg.
  • High CPU usage of 80% and more, might be observed when deploying the flows for this specified load. After the flows are deployed, the CPU usage will drop to the normal operating range.
  • Windows platform The overall ACMfg performance in a Window environment can be affected by the number of processes running. For an optimal performance, ensure that no other processes are running or consumes high CPU usage.

See, IBM App Connect for Manufacturing V2.0 performance reports.

OPC UA Server or datasource

It is recommended to use only one OPC UA Server or data source per message flow.

Deduplication mechanisms

Deduplication mechanisms based on OPCUA sequence numbers can be employed in the downstream nodes if an automatic failover happens in a high availability environment to avoid getting duplicate sequence numbers in the output.