Skip to main content

16.6 Replication & Delta Engineering

1. Replication flow role​

SAP Datasphere replication flows copy supported source assets to targets and support delta loads for applicable scenarios.

Current SAP documentation also exposes controls for source and target parallelism and other runtime settings.

2. Architecture​

                 SOURCE
                   |
          +--------+--------+
          |                 |
      Initial             Delta
          |                 |
          +--------+--------+
                   |
             Replication
                 Flow
                   |
                   v
                Target
                   |
                   v
             Harmonization

3. Initial load strategy​

For large S/4HANA datasets, consider:

  • source workload,
  • partitioning,
  • thread limits,
  • target write capacity,
  • scheduling,
  • business freeze requirements,
  • reconciliation.

Do not maximize parallelism blindly.

4. Delta strategy​

Document:

  • delta source,
  • change semantics,
  • deletion behavior,
  • package behavior,
  • restart behavior,
  • retention or reinitialization requirements,
  • reconciliation.

5. Parallelism​

Current Datasphere replication-flow documentation exposes source and target thread limits and maximum parallel delta transfers for supported scenarios.

Architectural rule:

Increase parallelism only after measuring the bottleneck.

If S/4HANA is the bottleneck, more Datasphere threads can make the situation worse.

6. Schema changes​

Source schema evolution requires explicit governance.

Where supported, Datasphere provides mechanisms for detecting compatible source schema changes for certain ABAP replication scenarios.

Never assume every source change is automatically safe.

7. Delete handling​

A delete is not the same as "no new row."

The target must know whether the source semantics represent:

  • insert,
  • update,
  • delete,
  • replacement,
  • reinitialization.

Validate delete behavior explicitly.

8. Reconciliation​

For a critical replication, reconcile:

  • record counts,
  • control totals,
  • key ranges,
  • changed-record counts,
  • latest timestamps,
  • business totals.

9. Failure recovery​

A production runbook should answer:

  • can the run be restarted?
  • from where?
  • is the target idempotent?
  • does a failed package need reprocessing?
  • when is reinitialization required?

Architect exercise​

A 2 TB S/4HANA source takes 14 hours for initial replication.

Design a performance investigation.

Identify:

  • source bottleneck,
  • network bottleneck,
  • source threads,
  • target threads,
  • partitioning,
  • scheduling,
  • reconciliation,
  • acceptable migration window.