Cloud Platforms & Apps: AWS, Azure, GCP, SaaS, PaaS, IaaS

Key Concepts and Cloud Service Models

1. Cloud Service Models: IaaS, PaaS, SaaS

Cloud computing delivers services in three primary models, each abstracting different layers of infrastructure. The National Institute of Standards and Technology (NIST) defines these as:

sequenceDiagram
    participant User
    participant SaaS_App as SaaS Application (e.g., Gmail)
    participant PaaS_Platform as PaaS Platform (e.g., Heroku)
    participant IaaS_VM as IaaS VM (e.g., AWS EC2)
    participant Cloud_Infrastructure as Cloud Provider

    User->>SaaS_App: Logs in (No control over backend)
    SaaS_App-->>User: Delivers service
    
    User->>PaaS_Platform: Deploys app (Controls app config)
    PaaS_Platform->>Cloud_Infrastructure: Uses underlying VMs
    
    User->>IaaS_VM: Installs OS/apps (Full control)
    IaaS_VM->>Cloud_Infrastructure: Runs on virtualized hardware

2. Major Cloud Platforms: AWS, Azure, Google Cloud

Amazon Web Services (AWS)

• Market Leader: ~33% global market share (2023).
• Key Services:
  • Compute: EC2 (IaaS), Lambda (serverless).
  • Storage: S3 (object storage), EBS (block storage).
  • Databases: RDS (managed SQL), DynamoDB (NoSQL).
  • AI/ML: SageMaker, Rekognition.
• Use Case: eSewa uses AWS for scalable payment processing during festivals (e.g., Dashain, Tihar), handling spikes in transactions via auto-scaling EC2 instances.

Microsoft Azure

• Enterprise Focus: Integrates with Windows Server, Active Directory, and Microsoft 365.
• Key Services:
  • Hybrid Cloud: Azure Arc for on-premises integration.
  • AI: Azure Cognitive Services (e.g., speech-to-text).
  • Databases: Cosmos DB (globally distributed NoSQL).
• Use Case: Nepal Rastra Bank (NRB) uses Azure for secure financial transaction monitoring, leveraging Azure Sentinel for threat detection.

Google Cloud Platform (GCP)

• Data and AI: Strong in machine learning (TensorFlow, Vertex AI).
• Key Services:
  • Compute: Compute Engine (IaaS), Cloud Functions (serverless).
  • Storage: Cloud Storage (object storage), Persistent Disk (block storage).
  • Networking: Global Load Balancing for low-latency apps.
• Use Case: Pathao uses GCP’s BigQuery for real-time ride analytics, optimizing driver routes and reducing wait times by 20%.

In the Real World

1. eSewa (Nepal)

   • Idea Used: Auto-scaling IaaS (AWS EC2).
   • How: During festivals, eSewa’s backend scales up EC2 instances dynamically to handle 10x transaction spikes. Without cloud, their on-prem servers would crash.

2. Daraz (Alibaba Group)

   • Idea Used: Serverless PaaS (AWS Lambda + API Gateway).
   • How: Daraz’s order processing uses Lambda to trigger workflows (e.g., inventory updates, shipping notifications) without managing servers. This reduces costs by 40% vs. traditional VMs.

3. NTC (Nepal Telecom)

   • Idea Used: SaaS for Customer Support (Zendesk).
   • How: NTC’s customer service team uses Zendesk (SaaS) to manage complaints, route tickets, and track resolutions—no IT overhead for software updates or hardware.

How Cloud Platforms Work: Deployment Models

Cloud services can be deployed in four models, each balancing control, cost, and scalability:

flowchart TD
    A["Public Cloud"] -->|"Example: AWS, GCP"| B["Multi-tenant, shared infrastructure"]
    C["Private Cloud"] -->|"Example: On-prem Azure Stack"| D["Single-tenant, dedicated resources"]
    E["Hybrid Cloud"] -->|"Example: NRB + Azure"| F["Mix of public/private for security/compliance"]
    G["Community Cloud"] -->|"Example: Healthcare consortium"| H["Shared by specific groups (e.g., banks)"]

Worked Example: Kathmandu Traffic Management

• Problem: Traffic jams during Dashain cause 3-hour delays.
• Solution: Hybrid Cloud (Azure + IoT):
  • Public Cloud (Azure IoT Hub): Collects real-time data from traffic cameras and GPS.
  • Private Cloud (On-prem): Processes sensitive data (e.g., license plate tracking) locally for privacy.
  • Outcome: Dynamic rerouting via SaaS apps (e.g., Pathao’s traffic API) reduces congestion by 25%.

Comparing Cloud Platforms: Feature Matrix

Security and Compliance in Cloud Platforms

Cloud providers offer shared responsibility models, where security duties split between the provider and the customer:

erDiagram
    CLOUD_PROVIDER ||--o{ CUSTOMER : "Hosts"
    CLOUD_PROVIDER {
        string provider_name PK
        string compliance_certifications
    }
    CUSTOMER ||--|{ APPLICATION : "Deploys"
    CUSTOMER {
        string customer_name PK
        string shared_responsibility_level
    }
    APPLICATION {
        string app_name PK
        string data_encryption
    }
    CLOUD_PROVIDER ||--o{ SECURITY_SERVICE : "Offers"
    SECURITY_SERVICE {
        string service_name PK
        string responsibility
    }

stateDiagram-v2
    [*] --> Provider: "Infrastructure (Physical Security, Network)"
    Provider --> Customer: "Client-Side Data, Encryption Keys"
    Customer --> [*]: "Application, OS, Data"

Real-World Example: NEPSE (Nepal Stock Exchange)

• Challenge: Secure trading platform with strict compliance (SEBI, Nepal Rastra Bank).
• Solution:
  • IaaS (Azure): Hosts VMs with hardened OS templates.
  • PaaS (Azure Key Vault): Manages encryption keys for sensitive data.
  • SaaS (Docusign): Uses for legally binding e-contracts.

Exam Tip

1. Memorize the 3 Service Models (IaaS/PaaS/SaaS):

   • IaaS: "I" for Infrastructure → VMs, storage.
   • PaaS: "P" for Platform → Runtime, middleware.
   • SaaS: "S" for Software → Ready-to-use apps (e.g., Gmail).
   • Exam trick: Questions often ask to match a scenario to the correct model. Example:

     "A bank uses a cloud provider’s managed database without controlling the underlying servers. Which model?" → PaaS.

2. Platform Comparisons:

   • AWS = Most services (best for startups needing flexibility).
   • Azure = Best for Microsoft ecosystems (e.g., Windows Server, Office 365).
   • GCP = Best for AI/ML and data (e.g., TensorFlow, BigQuery).
   • Exam tip: Always compare 2 platforms in answers (e.g., "AWS excels in compute, while Azure integrates better with Active Directory").

3. Deployment Models:

   • Public Cloud: Cheapest, least control.
   • Private Cloud: Most secure, highest cost (e.g., government).
   • Hybrid: Most common in enterprises (e.g., NTC + Azure).
   • Exam question: "Which model would you recommend for a hospital’s patient records?" → Private Cloud (HIPAA compliance).

4. Real-World Scenarios:

   • eSewa: IaaS (auto-scaling).
   • Pathao: PaaS (serverless functions).
   • NEPSE: Hybrid (public for trading, private for compliance).
   • Tip: Always tie examples to Nepal (e.g., Daraz, Khalti, NTC) to score extra marks.

5. Diagrams:

   • Draw layered models (e.g., IaaS/PaaS/SaaS stack).
   • Sketch deployment models (public/private/hybrid).
   • Use sequence diagrams for cloud workflows (e.g., how AWS Lambda triggers S3 events).

In the real world

• eSewa (AWS IaaS): Uses auto-scaling EC2 instances during festivals (Dashain/Tihar) to handle 10x transaction spikes without manual server upgrades. Real-world impact: Prevents system crashes during peak load (e.g., 2023 Dashain saw 5M+ transactions/day).
• Pathao (GCP PaaS): Leverages Google Cloud Functions (serverless) for real-time ride analytics, reducing driver wait times by 20% via optimized routing. Real-world impact: Saves 100K+ hours/year for drivers.
• NTC (SaaS): Uses Zendesk (SaaS) for customer support, cutting IT overhead by 30% (no need to maintain servers/software). Real-world impact: Resolves 90% of complaints within 24 hours.