How would you migrate 1TB of data using ADF?

Architectural logic: 1TB is large enough that naive single-threaded copy will miss SLAs; parallelism and staging are critical. Why parallel copy: ADF's Copy activity supports parallelCopies (default 1, max ~32 on Azure IR)—each copy runs concurrently, bounded by source/sink throughput and IR capacity. For databases, use physical partition option so ADF splits by partition key. For file-based (S3/Blob), use wildcard paths and binary copy to avoid schema parsing. Scalability: parallelCopies × dataIntegrationUnits determines throughput; 8 × 4 DIUs can yield 200–400 MB/s depending on source. Cost: Azure IR is pay-per-use; Self-Hosted IR adds VM cost. For 1TB at 100 MB/s ≈ 3 hours; at $0.002/DIU-hour, compute cost is minimal—network and staging storage dominate. Staged copy: If source has limited egress (e.g., on-prem), stage to Blob first with SHIR, then Azure IR copies to final sink—avoids saturating on-prem links. Run during off-peak; set up alerts; validate with checksums post-migration. For 1TB, expect 4–12 hours depending on network and source type.