13. Whiteboard: hybrid cloud and migration

Nobody migrates for fun. Draw the end state first, then the path, then the risk controls.

Whiteboard outcome

Today-state on the left (NAS, tape, S3/Azure/GCS/Wasabi), end-state on the right (B2 primary plus on-prem cache), and three migration lanes in the middle with TB and days written on each.

Draw it in 9 steps

1. Left column, “Today”: boxes for each source with TB: “AWS S3: 600 TB”, “NAS: 200 TB”, “LTO: 900 tapes”.
2. Right column, “End state”: “Backblaze B2” big box; small “on-prem NAS cache” box; arrows from apps.
3. Lane 1 (cloud to cloud): arrow S3 → B2 labeled “Universal Data Migration: Backblaze helps cover egress and transfer with qualifying commit; MD5 verified”.
4. Lane 2 (NAS): arrow NAS → B2 labeled “rclone / NAS Cloud Sync / backup software; bandwidth = X Gbps → Y days”.
5. Lane 3 (offline): truck icon NAS/tape → “Fireball (96 TB Synology, 10GbE)” → B2. Say: “When the wire is too slow, ship it.”
6. Under lanes: “Dual-write window” bracket. Say: “New writes go to both for N days, then we cut reads over.”
7. Verification box: “Inventory diff: object count, bytes, MD5/SHA1 sample”.
8. Cutover box: “Flip endpoint + keys in app config. Keep source read-only 30 days.”
9. Write the transfer formula: days = TB × 8,000 / (Gbps × 86,400 × efficiency). With 1 Gbps at 80%: 100 TB ≈ 11.6 days.

flowchart LR
  subgraph Today
    S3[AWS S3 / Azure / GCS / Wasabi]
    NAS[On-prem NAS / SAN]
    LTO[LTO tape]
  end
  subgraph End[End state]
    B2[(Backblaze B2)]
    CACHE[On-prem cache / NAS]
    APPS[Apps & backup software]
  end
  S3 -- UDM: subsidized, MD5-verified --> B2
  NAS -- rclone / Cloud Sync / backup sw --> B2
  NAS -- Fireball 96 TB --> B2
  LTO -- UDM tape services --> B2
  APPS --> CACHE --> B2
  APPS --> B2

Transfer time table (80 percent link efficiency)

Formula: seconds = bytes×8 / (bps × 0.8). 1 PB = 8×10^15 bits; at 10 Gbps × 0.8 = 8×10^9 bps → 10^6 s ≈ 11.6 days.

Full playbook in 19-migration-playbook.md.