On-Site vs. Off-Site vs. Cloud Backup Storage
| Attribute | On-Site Backup | Off-Site Backup | Cloud Backup |
|---|---|---|---|
| Physical Location | Same building as production systems | Geographically separate facility (different city or region) | Third-party cloud provider's data centers |
| Restore Speed | Fastest — LAN speeds; no WAN dependency | Slower — WAN or internet link required | Slowest for large data — limited by internet bandwidth |
| Cost | Lower — no recurring storage facility fees | Medium — facility fees or tape transport costs | Pay-per-GB; scales with data volume |
| Disaster Survival | Vulnerable to site-level disaster (fire, flood) | Protected from site disasters — different location | Protected from site disasters; provider-managed infrastructure |
| Encryption Requirement | Recommended — insider threat, media theft | Required — media transported outside facility control | Essentially mandatory — data in third-party environment |
| Best For | Fast everyday restores (file recovery, DB restore) | Long-term retention, disaster recovery | Unlimited scale, geographic redundancy, no media management |
Backup Frequency vs. Recovery Point Objective (RPO)
| Backup Interval | Maximum Data Loss (RPO) | Storage Impact | Typical Use Case |
|---|---|---|---|
| Continuous (Replication) | Seconds to minutes | Highest — ongoing bandwidth and storage | Financial transactions, critical databases, hot site DR |
| Hourly | Up to 59 minutes | High — 24 backups per day per system | High-change databases, transactional systems |
| Daily | Up to 23 hours 59 minutes | Medium — 30 backups per month | Most production servers, file servers, application data |
| Weekly | Up to 6 days 23 hours | Low — 4 backups per month | Slowly changing servers, reference data, archive systems |
| Monthly | Up to 30 days | Very low — 12 backups per year | Long-term retention, compliance archives |
Snapshot Incremental Design — What Each Day Captures
| Day | Total Data on System | Snapshot Captures | Storage Used by Snapshot | Can Restore To This Day? |
|---|---|---|---|---|
| Monday (Initial) | 100 GB | Full system state — all 100 GB | 100 GB (full copy) | Yes |
| Tuesday | 100 GB (40 GB changed) | Only the 40 GB that changed since Monday; 60 GB unchanged is not re-copied | 40 GB (delta only) | Yes — apply Monday + Tuesday delta |
| Wednesday | 120 GB (20 GB new data added) | Only the 20 GB new data added since Tuesday | 20 GB (delta only) | Yes — apply Monday + Tuesday + Wednesday deltas |
| Without Snapshots (Full Daily) | 120 GB | Full copy every day regardless of changes | 100 + 100 + 120 = 320 GB total | Yes — but 320 GB used vs. 160 GB for incremental |
Replication vs. Periodic Backup — Critical Differences
| Attribute | Replication | Periodic Backup |
|---|---|---|
| Timing | Continuous — near-real-time synchronization | Scheduled intervals — hourly, daily, weekly |
| Data Currency | Always current — seconds or minutes behind primary | Point-in-time — reflects state at backup job runtime |
| Ransomware Response | Dangerous — replicates encrypted files over good copies within seconds | Safe — historical snapshots capture state before encryption |
| Accidental Deletion Recovery | Difficult — deletion propagates to replicated site immediately | Easy — restore from backup taken before deletion |
| Infrastructure Failure Recovery | Excellent — replicated site has current data, activates quickly | Slower — restore from last backup; changes since backup are lost |
| Hot Site Support | Ideal — hot site constantly receives current data | Warm/cold site — must restore from backup on activation |
| Replaces Periodic Backup? | No — replication cannot recover from ransomware or accidental deletions | No — periodic backup alone cannot provide the fast recovery of replication |
Journaling — Failure Scenario Outcomes
| Scenario | What Happens to Storage Data | What Happens at Next Startup | Need to Restore from Backup? |
|---|---|---|---|
| No Journaling — Power fails mid-write | Partially written — data structures inconsistent; file size may not match actual data; database records half-updated | Storage appears corrupt; application may not open files; database may refuse to start | Yes — must restore from last good backup; all changes since backup are lost |
| With Journaling — Power fails during journal write | Nothing written to actual storage target yet — storage is in exactly the pre-write state | Incomplete journal entry discarded; storage remains consistent; no corruption | No — storage is clean; only the in-flight operation is lost (not committed) |
| With Journaling — Power fails during storage write (after journal) | Partial write to storage — but journal entry records what the completed operation should look like | System replays the journal entry to complete the interrupted write; data brought to consistent state before applications start | No — journal-based recovery completes automatically; no backup restore needed |
| With Journaling — Both writes complete successfully | Data fully and correctly written to storage | Journal entry cleared; ready for next operation | No — normal operation |
The Six Backup Planning Dimensions
| Dimension | Decision | Why It Matters | Common Options |
|---|---|---|---|
| 1. Data Volume | How much data must be backed up? | Determines media selection, transfer time, storage cost, and whether full daily backups are practical | Megabytes to petabytes; segment by system type |
| 2. Backup Type | Full, incremental, differential, or snapshot? | Determines storage space per backup job and restore complexity | Full (complete copy); Incremental (changes since last backup); Snapshot (VM/cloud point-in-time) |
| 3. Backup Media | Where is the backup physically stored? | Affects durability, access speed, cost, and off-site transport capability | Tape (long-term, portable); hard drive / NAS (fast, local); cloud (scalable, off-site) |
| 4. Storage Location | On-site, off-site, cloud, or hybrid? | Determines disaster resilience — on-site is fast; off-site survives site disasters | Most organizations use both on-site and off-site |
| 5. Software | What creates and restores the backups? | Must support all data types, media, and restore workflows; recovery software must be accessible even when the primary system is down | Third-party enterprise tools; OS built-ins; cloud-native backup services |
| 6. Schedule | How often — and when — do backup jobs run? | Determines RPO (data loss tolerance) and must balance backup window against production performance impact | Hourly, daily, weekly, monthly; staggered across different systems |