Scalability

This section provides best practices for designing orchestration workflows to maximize performance and scalability.

An application that performs large-scale data transformations will not scale well if it is implemented using an orchestration. Instead of using a single orchestration to process a large number of items, break the work into smaller chunks or consider using another method to process the data. In general, use orchestrations only to enable collaboration between systems and users.
Use orchestration workflows to solve the following business problems:
- Collaboration with existing legacy systems: If a legacy system can receive Web service calls, it can be called from asynchronous orchestration workflows that are executed during transitions in an application workflow.
  Important: It is recommended that you use asynchronous (not synchronous) orchestration workflows to do this. See Usage for synchronous orchestration workflow limitations.
- Collaboration with SOA capable systems: Application workflows (also known as "human workflows") can easily be integrated with other SOA (service-oriented architecture) systems. These integrations use orchestration workflows initiated from the human workflow to call Web services for passing data to and receiving responses from the external systems. In addition, because the Event Manager can be called as a Web service, asynchronous orchestrations can be initiated from events occurring in the external systems to communicate back to the SBM application.
- Intelligent data enrichment: Users make decisions and take actions based on information available within the system. Often they can make better decisions if they have access to related information stored outside of the system. A synchronous orchestration workflow can implement business logic to collect and coordinate data from several sources to present in a form. For example, a form in a credit approval application can include a customer's credit information.
- Validation: Scripts can be used to validate input in on-premise environments, but are not available in on-demand environments. Synchronous orchestration workflows that use programming constructs such as comparisons, string and number manipulation, and loops can be used to validate data that users entered. Validation can also occur through the use of JavaScript in custom forms.
- Process-level synchronization: In this scenario, the data between two or more systems is continuously synchronized as items flow through various business processes. Another synchronization scenario is not suitable for orchestration workflows and is described below.
Do not use orchestration workflows to perform the following tasks:
- Visual programming: SBM Composer provides the ability to visually create orchestration workflows using constructs such as loops, decisions, data manipulation, and fault handling. While it is possible to use these constructs as a general purpose programming language to create complex procedural programs, it is discouraged. It is better to use these constructs to apply business logic while coordinating or orchestrating the Web services called in the workflow.
  Drawbacks of adding long modules within orchestration workflows include: orchestration workflows become too long, so they are difficult to view, debug, and maintain; designers lose the benefit of compile time checks that other languages such as Java offer; and performance of these constructs is nowhere near as fast as that of traditional programming languages. Complex logic or data manipulation should be moved out of orchestration workflows and into Web services that can be invoked as a part of the orchestration workflow.
- ETL processing: Data warehousing Extract, Transform, and Load (ETL) processing is data intensive and requires minimum overhead for each record and very fast data transformation processing. Implementing ETL with an orchestration workflow entails at least two Web service calls for each record that is processed, as well as the overhead for each record for data transformation. Using an orchestration workflow for ETL processing results in poor performance and unacceptably complex visual programming.
- Batch process synchronization: In this scenario, synchronization takes place after the fact in a batch process. It involves processing a large number of records in multiple systems to ensure that they remain synchronized after changes to one or more system. An example of this is synchronizing a customer or product master database with data stored in distributed systems. To achieve this using an orchestration workflow, you would need to iterate through each record on each system, making multiple Web service calls for each iteration; comparing the data between systems, and if necessary, update the data by making additional Web service calls.