The Inspiration: Partitioned Pipeline Orchestration
In enterprise ETL/ELT pipeline management (e.g., Informatica, legacy ODI), targeted data recovery is a fundamental capability. When source data is corrected in an upstream ERP/HCM system, pipeline operators re-run only the affected partition. FDI needs this same Selective Reprocessing control within its Data Pipeline management console.
Problem Statement
Currently, when a data correction occurs in Oracle Fusion Cloud (eg. GL adjustment, or a recalculated payroll):
- Excessive Resource Consumption: To fix a data gap in a single Legal Entity or date range, admins often have to trigger a Reload. This re-extracts millions of unchanged rows, spiking ADW compute.
- Extended Downtime: A full reset can take hours to complete. During this window, business users experience stale data, data inconsistencies, or degraded report performance across the entire functional area.
Proposed Solution: Selective Pipeline Reload Controls
Introduce a Granular Pipeline Reprocessing utility within the FDI Console for pipeline data loads:
- Targeted Filter Controls: Allow administrators to trigger an ad-hoc pipeline refresh using specific filter criteria:
- Date Range / Period: (e.g., Reload only Q2 2026 or 15-SEP-2026 to 20-SEP-2026)
- Organizational Boundary: (e.g., Reload only Legal Entity UK or US Operations)
- Source Batch / Extract ID: Re-run a specific failed extraction slice.
- Delta Merge Logic: Ensure the pipeline target tables in ADW selectively delete and replace only the matching records, leaving all other historical data intact.
- Execution Preview: Display an estimate of impacted record counts and processing time before launching the targeted reload.
Business Value
- 80%+ Faster Recovery: Cuts pipeline recovery times from hours to minutes by focusing processing power strictly on the affected data slice.
- Significant ADW Cost Savings: Reduces unnecessary database compute consumption by eliminating redundant global re-extracts.
- High System Availability: Keeps reports and workbooks available for business users while localised data updates run silently in the background.