Backup & Software Repository Services: Types, Tools & Strategies

TAKEAWAYS:

• Backup types differ by scope (full, incremental, differential) and must align with RPO/RTO goals—e.g., Ncell’s daily incremental backups of SIM card data ensure 1-hour recovery.
• Software repositories (YUM, APT, Docker Hub) automate package management, reducing manual updates (e.g., Daraz’s Linux servers use APT for real-time order-processing dependencies).
• Disaster recovery requires tiered backups (local + cloud) and tested restoration—like NTC’s fiber-optic route backups to prevent outages during landslides.
• Linux/Windows commands (tar, robocopy, apt-cache) verify backups and repositories; master these for exam scenarios (e.g., "Check if a backup file is corrupted").
• Repository security uses GPG keys (e.g., WhatsApp’s code-signing repos) and access controls to prevent tampering.
• Cloud vs. on-premise backups trade cost for latency—e.g., eSewa uses AWS S3 for offsite backups but keeps critical transaction logs on-premise for compliance.

---

1. Backup Services: Types, Strategies, and Tools

1.1 What is Backup?

Backup is the process of copying and archiving data to a secondary storage medium to prevent loss from hardware failure, cyberattacks, or human error. Key terms:

• RPO (Recovery Point Objective): Maximum acceptable data loss (e.g., 1 hour for Ncell’s billing system).
• RTO (Recovery Time Objective): Maximum downtime tolerated (e.g., 4 hours for Daraz’s order processing).

1.2 Types of Backups

Worked Example: Ncell’s SIM Card Database

• RPO: 15 minutes (critical for subscriber data).
• Strategy:
  • Full backup: Weekly (stored offline).
  • Incremental: Every 15 minutes (stored on NAS + encrypted cloud).
• Recovery:
  • If corruption occurs at 3:45 PM, restore the full backup (Sunday) + incremental from 3:30 PM.

1.3 Backup Tools and Commands

Linux:

# Create a full backup (tar + compression)
tar -czvf backup_2024.tar.gz /var/www/html

# Verify checksum (MD5)
md5sum backup_2024.tar.gz

# Schedule daily incremental backups (cron)
0 2 * * * tar -czvf /backups/incremental_$(date +\%Y\%m\%d).tar.gz /var/log/

Windows:

# Create a full backup (robocopy)
robocopy C:\Data E:\Backup\Full\ /MIR /Z /R:3 /W:5

# Verify backup integrity
Get-FileHash -Algorithm SHA256 E:\Backup\Full\*.zip

Cloud Tools:

• AWS Backup: Automates cross-region replication (e.g., eSewa’s transaction logs).
• Veeam: Agentless backups for VMs (used by NTC’s core routers).

---

2. Software Repository Services

2.1 What is a Repository?

A software repository is a centralized storage for software packages, updates, and dependencies. It enables:

• Version control (e.g., rolling back a misconfigured update).
• Dependency resolution (e.g., installing nginx pulls in libssl).
• Security patching (e.g., WhatsApp’s auto-updates from Docker Hub).

2.2 Types of Repositories

Worked Example: Daraz’s Order Processing System

• Repository: APT (Ubuntu 22.04).
• Dependency Chain:

  apt install python3-pip       # Installs Python + pip
  pip install flask==2.0.1      # Locks version for reproducibility
  

• Why? Ensures all servers run the same flask version to avoid crashes during Black Friday sales.

2.3 Repository Security

• GPG Keys: Sign packages to prevent tampering (e.g., WhatsApp’s .deb files).

  gpg --import whatsapp_repo.gpg
  

• Access Control: Restrict repositories to specific IPs (e.g., Ncell’s internal Docker Hub).
• Vulnerability Scanning: Tools like apt list --upgradable or docker scan.

---

3. Disaster Recovery Planning (DRP)

3.1 DRP vs. Backup

3.2 DRP Components

1. Tiered Backups:
   • Local: NAS (fast, but vulnerable to fire).
   • Offsite: Cloud (AWS S3, encrypted).
   • Air-Gapped: Tape drives (for ransomware).
2. Failover Testing:
   • Simulate a power outage (e.g., NTC’s UPS + generator test every quarter).
3. Documentation:
   • Runbook: Step-by-step recovery (e.g., "Replace failed router with spare from rack B").

Worked Example: Kathmandu Traffic Outage

• Scenario: Major fiber cut during Dashain.
• DRP Steps:
  1. Activate backup route via NTC’s redundant fiber ring.
  2. Restore traffic data from last incremental backup (RPO: 30 mins).
  3. Notify users via SMS (using Khalti’s bulk SMS API).

---

4. Exam Tip

What Examiners Want to See:

1. Backup Strategies:
   • Always compare full vs. incremental vs. differential with RPO/RTO examples.
   • Trace a recovery scenario (e.g., "If a backup fails at 3:00 PM, how do you restore?").
2. Repository Commands:
   • Linux: apt, yum, docker pull.
   • Windows: robocopy, Get-FileHash.
   • Security: Mention GPG keys or access controls.
3. DRP:
   • Tiered backups (local + cloud + air-gapped).
   • Real-world tie-in: NTC, eSewa, or Daraz examples.

Common Pitfalls:

• Forgetting to verify backups (always include md5sum or Test-Restore in answers).
• Confusing incremental vs. differential (draw a timeline in exams).
• Ignoring security (e.g., encrypted backups, GPG-signed repos).

---

In the Real World

1. eSewa’s Transaction Logs:

   • Backup Strategy: Full backup weekly + incremental every 15 minutes (RPO: 15 mins).
   • Repository: Private Docker Hub for microservices (e.g., esewa-payment-service:v2.1).
   • DRP: If the primary database fails, restore from AWS S3 + failover to a hot standby server in Pokhara.

2. Ncell’s SIM Card Database:

   • Backup: Incremental backups every 15 mins (stored on NAS + encrypted cloud).
   • Repository: YUM for OS updates (e.g., yum update kernel).
   • DRP: If a data center floods, switch to a pre-configured backup server in Chitwan.

3. Daraz’s Order Processing:

   • Backup: Differential backups nightly (RPO: 1 hour).
   • Repository: APT for dependencies (e.g., apt install python3.8).
   • DRP: During Diwali sales, if a server crashes, spin up a new VM from a snapshot in AWS.

---

Practice Questions (Exam-Style)

1. Short Answer:

   • "Explain how Ncell would implement a differential backup for its SIM card database with an RPO of 1 hour." (Hint: Full backup Sunday, differentials Mon-Sat.)

2. Command-Based:

   • "Write Linux commands to: a) Create a full backup of /var/www/html to /backups/. b) Verify the backup’s integrity using checksums. c) Schedule a daily incremental backup at 2 AM." (Hint: tar, md5sum, cron.)

3. Scenario:

   • "The primary database of a bank (like NMB) is corrupted. The last full backup was 3 days ago, and incremental backups are taken daily. Describe the recovery steps if the RTO is 4 hours." (Hint: Restore full + last 3 incrementals.)