– Pre-Installation Planning Guide –

Follow these steps for any production scenario involving Data Hub. Some are specific to the ERP integration.

You must consider the guidelines under this heading and the General Guidelines heading as part of your go-live strategic plan. Any Data Hub deployment that has not strictly followed all of these guidelines may experience performance issues and is considered to be misconfigured.

SAP IDoc Guidelines

Send appropriate data from ECC - Every IDoc sent to Data Hub results in many RawItems being created. More specifically, each segment of an IDoc becomes one Raw Item. Data Hub performance degrades with every additional unneeded Raw Item. This degradation is non-linear, so operating on large sets of Raw Items can have a negative impact on performance. To maximize performance of Data Hub, carefully analyze the specific ECC instance and what data is contained in each IDoc sent. Sending the minimum number of IDocs, and the relevant (minimum possible) segments inside of each IDoc has been shown to give massive performance gains. Consider investigating the following three transactions to improve the performance:

  1. BD53 - Reduction of Message Types

  2. BD56 - Maintain IDoc Segment Filters

  3. BD64 - Maintenance of Distribution Model

IDoc Filtering - The sapidocintegration extension filters out (removes) duplicates based on their unmapped attributes. As a result, it creates unique rows and reduces the total amount of raw items input to Data Hub. It uses a conservative approach to only remove consecutive duplicates. It reads in each of the relevant Data Hub raw model types in advance to validate the required set. Then it only creates those raw items that have been defined in the raw model.

General Guidelines

Database Volume - Data Hub performance degrades with increased volumes of items in the database. The default behavior of Data Hub is to maintain all items in the database for a comprehensive history and auditing perspective. Continually clean up this data to ensure that Data Hub provides consistent performance. For more information, see Activating Data Hub Database Cleanup.

Operation Queuing - Data Hub operates by loading items into a specific pool, and each pool operates independently from other pools. Composition and Publication actions occur concurrently. Every time either of these actions is initiated, whether manually or through an event-driven process, it is queued in a single queue.

Composition actions and publication actions each have their own queue, and all actions are processed in order in the queues. The status of each action shows as 'PENDING' with a count of 0 until it begins to process. If you are using a datahub.max.composition.action.size of 100,000, for example, when those 100,000 items are composed, composition continues with the next action, and so on. Publication can be initiated at any time and runs simultaneously with composition.

Tuning Steps

Procedure

  1. Database tuning:

    • For Oracle, review the execution plan to optimize performance. The execution plans may be different by customer installation.

    • Index fragmentation due to the heavy read/write activity can be an issue and has been observed in MSSQL.

  2. If Auto-Compose and Auto-Publish are used, adjust the sleep times against the data gathered to avoid unnecessary queuing of additional actions.
Related Information