108  Cloud Deployment Models for IoT

In 60 Seconds

Four cloud deployment models exist: public (shared, pay-per-use), private (dedicated, full control), hybrid (recommended for enterprise IoT), and community (shared among similar organizations). Use edge for real-time decisions under 100ms and cloud for historical analytics – most IoT systems need both tiers.

Minimum Viable Understanding

• Four deployment models exist: public cloud (shared, pay-per-use), private cloud (dedicated, full control), hybrid cloud (combines both), and community cloud (shared among similar organizations).
• Hybrid cloud is recommended for most enterprise IoT: keep sensitive data (PHI, trade secrets) on private infrastructure while using public cloud for analytics, ML, and scalable storage.
• Edge vs. cloud is not either/or: use edge for real-time decisions under 100ms, cloud for historical analytics and ML training – most systems need both.

Sensor Squad: Where Should Our Data Live?

The Sensor Squad had a big argument about where to store all their readings.

“Put everything in the PUBLIC cloud!” said Sammy the Sensor. “It’s like renting a storage unit – cheap, lots of space, and someone else takes care of it!”

“NO WAY!” shouted Bella the Battery. “Some of our data is TOP SECRET – like hospital patient readings. That needs a PRIVATE cloud – our own locked room that only we control!”

Max the Microcontroller had the best idea: “Why not BOTH? It’s called a HYBRID cloud!”

“Here’s how it works,” Max explained. “Sensitive patient data stays in our private cloud – safe and under our control. But when we want to run fancy machine learning to find patterns in years of data, we send ANONYMIZED copies to the public cloud where there are super-powerful computers!”

Lila the LED added one more thing: “And for REALLY urgent stuff – like detecting a dangerous heartbeat – we don’t even send it to ANY cloud. We process it right here at the EDGE, next to the patient. Milliseconds matter when lives are at stake!”

The lesson: Different data needs different homes. Sensitive data gets a private room, big analysis gets the public playground, and urgent decisions happen right at the edge!

For Beginners: Cloud Deployment Models Explained

When people say “the cloud,” they mean computers in a data center that you access over the internet. But there are different ways to set this up:

• Public Cloud (AWS, Azure, Google Cloud): Like renting an apartment – shared building, someone else maintains it, you pay monthly. Great for flexibility but you share resources with others.
• Private Cloud: Like owning your own house – you control everything but you pay for all the upkeep. Best when you need maximum security or must follow strict regulations.
• Hybrid Cloud: Like having a house AND renting storage space – keep important things at home, use rented space for overflow. This is what most large organizations choose.

For IoT specifically, the choice depends on how sensitive your data is and how fast you need to act on it. A smart factory might process safety alerts locally (edge), store machine data privately (private cloud), and run predictive maintenance AI in the public cloud.

108.1 Learning Objectives

By the end of this chapter, you will be able to:

• Compare Deployment Models: Evaluate public, private, hybrid, and community clouds for IoT workloads
• Design Hybrid Architectures: Architect solutions that balance cloud processing with edge and on-premises requirements
• Address Data Sovereignty: Apply deployment models to meet regulatory and compliance requirements
• Justify Placement Decisions: Determine where to process data based on latency, cost, and security trade-offs

108.2 Prerequisites

Before diving into this chapter, you should be familiar with:

• Cloud Computing Fundamentals: Understanding of NIST cloud model
• Cloud Service Models: Knowledge of IaaS, PaaS, and SaaS

108.3 Deployment Models Overview

Figure 108.1: Cloud deployment models - public, private, hybrid, and community cloud configurations

Figure 108.2: Cloud deployment models comparison.

108.4 Public Cloud

Owned and operated by third-party provider, shared among multiple customers.

Examples: AWS, Google Cloud, Microsoft Azure, IBM Cloud

IoT Use Cases:

• Rapid prototyping and development
• Startups with limited capital
• Variable workloads
• Global IoT deployments

Advantages:

• Zero upfront infrastructure cost
• Rapid scalability
• Global availability
• Managed services

Disadvantages:

• Less control over security
• Vendor lock-in concerns
• Compliance challenges for sensitive data

Figure 108.3: Public vs Private Cloud deployment comparison

108.5 Private Cloud

Dedicated infrastructure for single organization, on-premises or hosted.

Examples: OpenStack, VMware vCloud, Microsoft Azure Stack

IoT Use Cases:

• Healthcare IoT (HIPAA compliance)
• Industrial IoT with proprietary processes
• Government/military applications
• High-security requirements

Advantages:

• Complete control over infrastructure and data
• Customization to specific requirements
• Enhanced security and privacy
• Regulatory compliance easier

Disadvantages:

• High capital expenditure
• Requires IT staff for management
• Limited scalability compared to public cloud
• Longer provisioning times

108.6 Public Cloud vs Private Cloud Trade-off

Tradeoff: Public Cloud vs Private Cloud for IoT Data

Option A: Public Cloud (AWS, Azure, GCP) - Multi-tenant shared infrastructure with pay-per-use pricing and unlimited scalability.

Option B: Private Cloud (On-premises, OpenStack) - Dedicated single-tenant infrastructure with full control and enhanced security.

Decision Factors:

• Choose Public Cloud when: Rapid scaling is needed, capital budget is limited, workloads are variable, global distribution is required, or time-to-market is critical. Best for startups, prototypes, and non-sensitive IoT data.

• Choose Private Cloud when: Regulatory compliance requires data residency (HIPAA, GDPR), intellectual property must stay on-premises, latency requirements demand local processing, or predictable high-volume workloads make owned infrastructure cheaper over 3+ years.

• Consider Hybrid when: Sensitive data (PHI, trade secrets) needs private storage while analytics and ML benefit from public cloud scalability. The orchestration complexity adds 20-30% operational overhead but provides best-of-both-worlds flexibility.

108.7 Hybrid Cloud

Combination of public and private clouds with orchestration between them.

Figure 108.4: Hybrid Cloud IoT Architecture: Data Governance Across Private and Public Clouds

108.7.1 Real-World Hybrid IoT Example: Hospital Patient Monitoring

Scenario: 500-bed hospital with 2,000 patient monitoring devices.

Data Classification:

Data Type	Volume	Latency Req	Deployment
Real-time vitals (PHI)	2K devices x 1 msg/sec	<1 sec	Private Cloud (HIPAA)
Historical trends (anonymized)	100 GB/day	Minutes OK	Public Cloud (cheap storage)
ML training data (de-identified)	10 TB dataset	Hours OK	Public Cloud (GPU clusters)
Alerts & alarms (critical)	50 events/day	<500 ms	Private Cloud (reliability)

Hybrid Architecture Decision:

Private Cloud (On-Premises):

• What: Dell VxRail hyper-converged infrastructure (200 TB, 50 VMs)
• Why: HIPAA requires PHI under direct control, low latency for critical alarms
• Cost: $250K CAPEX + $30K/year maintenance
• Pros: Full control, <1ms latency, meets compliance

Public Cloud (AWS):

• What: S3 Glacier (archives), SageMaker (ML training), Lambda (batch processing)
• Why: Elastic capacity for ML training, 90% cheaper storage for 7-year retention
• Cost: $5K/month ($60K/year OPEX)
• Pros: Infinite scalability, pay-per-use, access to advanced ML tools

Cost Comparison (5-year TCO):

Approach	5-Year Total	Notes
Public Cloud Only	$900K	Compliance audit costs +$50K/year
Private Cloud Only	$610K	Cannot scale for ML training
Hybrid (Recommended)	$910K	Best balance: compliance + scalability

Trade-off Analysis: Hybrid costs $10K more over 5 years but provides ML innovation, 10x scalability for pandemic response, and disaster recovery.

Putting Numbers to It

Bandwidth Cost Analysis: Edge Pre-Processing vs Raw Cloud Upload

A fleet of 1,000 autonomous delivery robots generates GPS coordinates (8 bytes lat/long), speed (2 bytes), battery level (2 bytes), and sensor status (2 bytes) every second. Should they send raw data to cloud or pre-process at edge?

Raw data upload (Option A):

Data per robot per second: \[D_{\text{raw}} = 8 + 2 + 2 + 2 = 14 \text{ bytes/reading}\]

Fleet-wide per day: \[D_{\text{daily}} = 1000 \times 14 \text{ bytes/sec} \times 86400 \text{ sec/day} = 1,209,600,000 \text{ bytes} \approx 1.21 \text{ GB/day}\]

Cellular data cost (assume \(\$1.00\)/MB for IoT plans): \[\text{Cost}_{\text{daily}} = 1210 \text{ MB} \times \$1.00 = \$1,210/\text{day}\] \[\text{Cost}_{\text{annual}} = \$1,210 \times 365 = \$441,650/\text{year}\]

Edge-aggregated upload (Option B):

Pre-process on robot’s edge compute: - Aggregate readings over 5-minute windows (300 seconds) - Send only: average speed, min/max battery, route summary, exception events

Aggregated data per robot per 5 minutes: \[D_{\text{agg}} = 20 \text{ bytes (summary)} + 14 \text{ bytes} \times 10 \text{ (exceptions)} = 160 \text{ bytes per 5-min}\]

Fleet-wide per day: \[D_{\text{daily}} = 1000 \times 160 \times \frac{1440}{5} = 46,080,000 \text{ bytes} \approx 46 \text{ MB/day}\]

Annual cost: \[\text{Cost}_{\text{annual}} = 46 \text{ MB/day} \times 365 \times \$1.00 = \$16,790/\text{year}\]

Savings: \((\$441,650 - \$16,790)/\$441,650 = 96.2\%\) reduction!

Key insight: Edge processing reduces bandwidth costs by \(300\times\) (raw 1.21 GB/day → aggregated 46 MB/day) while retaining critical information. The edge CPU cost (a few cents per device) is negligible compared to cellular data savings. This is why hybrid edge-cloud architectures dominate real-world IoT deployments.

108.8 Edge Processing vs Cloud Processing

Tradeoff: Edge Processing vs Cloud Processing for IoT Analytics

Option A: Edge Processing - Run analytics on gateways near IoT devices. Sub-100ms latency, operates during network outages, reduces bandwidth costs.

Option B: Cloud Processing - Send all data to cloud for centralized analytics. Unlimited compute for ML/AI, unified dashboards, simpler device firmware.

Decision Factors:

• Choose Edge when: Latency requirements are under 100ms (industrial control), bandwidth costs are significant (cellular IoT at $0.50-2.00/MB), network connectivity is unreliable, or data privacy requires local processing.

• Choose Cloud when: Advanced ML models require GPU clusters unavailable at edge, cross-device analytics need centralized data, historical analysis spans months/years, or edge hardware constraints limit processing capability.

• Bandwidth savings calculation: 1000 sensors at 1 reading/second = 86.4M readings/day. Raw to cloud: 8.64 GB/day. Edge aggregation (5-minute averages): 28.8 MB/day. Annual savings: $1,500-6,300 per 1000 sensors on cellular.

Figure 108.5: Cloud-Edge Data Flow Timeline showing latency at each stage.

108.9 IoT Scale Cost Comparison

Figure 108.6: IoT Scale Cost Comparison showing optimal deployment models at different scales.

108.10 Hospital Deployment Knowledge Check

108.11 Cloud Deployment Cost Calculator

Use this calculator to estimate and compare 5-year total cost of ownership for different deployment models based on your IoT system scale.

viewof numDevices = Inputs.range([100, 500000], {value: 1200, step: 100, label: "Number of IoT devices"})

viewof msgsPerDevPerDay = Inputs.range([10, 10000], {value: 288, step: 10, label: "Messages per device per day"})

{
  // AWS IoT Core pricing: $0.08 per million messages
  const awsPerMillionMsg = 0.08;
  const totalMsgsPerDay = numDevices * msgsPerDevPerDay;
  const totalMsgsPerYear = totalMsgsPerDay * 365;
  const publicCloudMsgCost = (totalMsgsPerYear / 1e6) * awsPerMillionMsg;
  const publicCloudComputeAnnual = Math.max(1200, numDevices * 0.002 * 12); // rough EC2 estimate
  const publicCloudTotalAnnual = publicCloudMsgCost + publicCloudComputeAnnual;
  const public5yr = publicCloudTotalAnnual * 5 + 8000; // setup

  // Private cloud: high upfront, lower per-year
  const privateCapex = Math.max(45000, numDevices * 30);
  const privateOpexAnnual = Math.max(30000, numDevices * 5);
  const private5yr = privateCapex + privateOpexAnnual * 5;

  // Hybrid: mid capex, lower opex than private
  const hybridCapex = Math.max(15000, numDevices * 10);
  const hybridOpexAnnual = Math.max(15000, numDevices * 3);
  const hybrid5yr = hybridCapex + hybridOpexAnnual * 5;

  const minCost = Math.min(public5yr, private5yr, hybrid5yr);
  const winner = public5yr === minCost ? "Public Cloud" : (private5yr === minCost ? "Private Cloud" : "Hybrid Cloud");

  function bar(cost, max) {
    const pct = (cost / max * 100).toFixed(0);
    return `<div style="background:#3498DB;height:16px;width:${pct}%;border-radius:3px;"></div>`;
  }
  const maxCost = Math.max(public5yr, private5yr, hybrid5yr);

  return html`<div style="background: var(--bs-light, #f8f9fa); padding: 1rem; border-radius: 8px; border-left: 4px solid #3498DB; margin-top: 0.5rem;">
    <p><strong>Scale:</strong> ${numDevices.toLocaleString()} devices × ${msgsPerDevPerDay.toLocaleString()} msgs/day = ${(totalMsgsPerYear/1e6).toFixed(1)}M messages/year</p>
    <table style="width:100%;border-collapse:collapse;font-size:0.9em;margin-top:0.5rem;">
      <thead><tr style="background:#2C3E50;color:white;"><th style="padding:6px;">Model</th><th style="padding:6px;">5-Year TCO</th><th style="padding:6px;">Relative Cost</th></tr></thead>
      <tbody>
        <tr><td style="padding:6px;">Public Cloud</td><td style="padding:6px;">$${public5yr.toLocaleString(undefined,{maximumFractionDigits:0})}</td><td style="padding:6px;">${bar(public5yr, maxCost)}</td></tr>
        <tr style="background:#f0f0f0;"><td style="padding:6px;">Private Cloud</td><td style="padding:6px;">$${private5yr.toLocaleString(undefined,{maximumFractionDigits:0})}</td><td style="padding:6px;">${bar(private5yr, maxCost)}</td></tr>
        <tr><td style="padding:6px;">Hybrid Cloud</td><td style="padding:6px;">$${hybrid5yr.toLocaleString(undefined,{maximumFractionDigits:0})}</td><td style="padding:6px;">${bar(hybrid5yr, maxCost)}</td></tr>
      </tbody>
    </table>
    <p style="margin-top:0.5rem;"><strong>Recommended:</strong> <span style="color:#16A085;font-weight:bold;">${winner}</span> (lowest 5-year TCO at this scale)</p>
    <p style="font-size:0.8em;color:#7F8C8D;">Note: Estimates are approximate. Private/hybrid cloud costs assume on-premises infrastructure and 0.25 FTE IT operations.</p>
  </div>`;
}

Interactive Quiz: Match Concepts

Interactive Quiz: Sequence the Steps

Key Concepts

• Public Cloud: Infrastructure owned and operated by a third-party provider (AWS, Azure, GCP) and shared among multiple tenants, offering the lowest cost and highest elasticity with reduced control over data residency
• Private Cloud: Dedicated cloud infrastructure operated exclusively for a single organization, either on-premises or hosted, providing maximum control and compliance capability at higher cost than public cloud
• Hybrid Cloud: An architecture combining public and private cloud resources with orchestration enabling workload portability, allowing sensitive IoT data to remain on-premises while burst workloads use public cloud
• Community Cloud: Shared infrastructure for organizations with common requirements (e.g., government agencies, healthcare networks) that split costs while maintaining sector-specific compliance and security controls
• Data Sovereignty: Legal requirements mandating that IoT data remain within specific geographic boundaries or jurisdictions, often driving private or hybrid cloud selection for regulated industries
• Cloud Bursting: The pattern where IoT workloads run on private infrastructure during normal load but automatically overflow to public cloud during peak demand, optimizing cost while maintaining baseline performance guarantees
• Multi-Cloud: Distributing IoT workloads across multiple cloud providers to avoid vendor lock-in, increase resilience, or access best-of-breed services, at the cost of significantly increased operational complexity

Common Pitfalls

1. Choosing Private Cloud for Cost Savings

Private cloud is typically 2–3× more expensive than public cloud when accounting for hardware, operations, and maintenance. It is justified by compliance, data sovereignty, or ultra-low-latency requirements — not cost. Use public cloud unless you have a specific regulatory or technical reason.

2. Underestimating Hybrid Cloud Complexity

Hybrid cloud requires consistent identity management, network connectivity (VPN/Direct Connect), synchronized orchestration, and unified monitoring across environments. Teams frequently underestimate the 6–12 months of integration work required before IoT workloads can span environments reliably.

3. Ignoring Data Egress Costs in Multi-Cloud

Cloud providers charge $0.08–0.12/GB for data transferred out of their networks. An IoT system generating 1 TB/day of cross-cloud traffic incurs $80–120/day ($29,000–44,000/year) in egress fees alone. Model data transfer costs before selecting multi-cloud architecture.

4. Assuming All Regions Have the Same Services

Deploying IoT workloads in a region that lacks required managed services (e.g., AWS IoT Core is unavailable in some regions). Always verify service availability in your target regions before designing the architecture.

Label the Diagram

💻 Code Challenge

108.12 Summary

This chapter covered cloud deployment models for IoT:

1. Public Cloud: Best for startups, variable workloads, rapid prototyping
2. Private Cloud: Best for compliance, security-sensitive applications
3. Hybrid Cloud: Best balance of compliance and scalability - recommended for most enterprise IoT
4. Edge-Cloud Integration: Use edge for real-time (<100ms), cloud for analytics and storage
5. Scale Considerations: Optimal deployment model changes with device count

108.13 Knowledge Check

108.14 Worked Example: Cloud Deployment Model Selection for a Smart Factory

Scenario: A German automotive parts manufacturer (500 employees, 3 factory floors) needs to deploy an IoT platform for 1,200 CNC machines and assembly robots. Requirements include real-time machine health monitoring, predictive maintenance using ML, compliance with EU data residency (GDPR), and integration with ERP systems.

Option A: Public Cloud Only (AWS IoT)

Component	Monthly Cost	Notes
AWS IoT Core (1,200 devices)	$1,800	$0.08/million messages
EC2 instances (ML training)	$3,200	p3.2xlarge for training
S3 storage (5 TB/month)	$115	Sensor time-series data
Data transfer out (500 GB)	$45	Dashboard, API responses
Monthly total	$5,160

Pros: No infrastructure to manage, elastic ML compute, 99.99% uptime SLA. Cons: Sensor data leaves factory premises (GDPR risk for employee proximity data), 20-50 ms latency to EU-West region (too slow for emergency stops), vendor lock-in to AWS APIs.

Option B: Private Cloud Only (On-Premises OpenStack)

Component	Monthly Cost (amortized over 5 years)	Notes
Server hardware (6 nodes)	$2,500	Dell PowerEdge cluster
Storage array (100 TB)	$800	NetApp or similar
Networking (10GbE switches)	$300	Spine-leaf fabric
IT staff (0.5 FTE)	$4,000	Operations and maintenance
Software licenses	$1,200	VMware, monitoring tools
Monthly total	$8,800

Pros: Full data sovereignty, <5 ms latency to factory floor, no recurring cloud fees after amortization. Cons: Limited ML compute (no GPU elastic scaling), capacity planning required, higher upfront capital ($180,000).

Option C: Hybrid Cloud (Recommended)

Component	Monthly Cost	Notes
On-premises edge cluster (3 nodes)	$1,200	Real-time control + data lake
AWS IoT Greengrass (local ML inference)	$150	Runs trained models on-prem
AWS SageMaker (ML training only)	$800	Train monthly, deploy locally
S3 for anonymized training data	$50	No PII leaves premises
VPN interconnect	$100	Site-to-site encrypted tunnel
IT staff (0.25 FTE)	$2,000	Less than full private cloud
Monthly total	$4,300

Why hybrid wins:

Requirement	Public	Private	Hybrid
Real-time control (<10 ms)	Fails (50 ms)	Passes	Passes (local)
GDPR data residency	Risk	Passes	Passes (PII stays local)
ML training compute	Passes	Limited	Passes (cloud GPU on-demand)
5-year TCO	$309,600	$528,000	$258,000
Upfront capital	$0	$180,000	$45,000
Operational complexity	Low	High	Medium

Decision: Hybrid cloud at $258,000 over 5 years saves 17% versus public-only and 51% versus private-only, while meeting all regulatory and latency requirements. The key architectural pattern: sensitive real-time data stays on-premises, while compute-intensive ML training uses elastic public cloud resources on anonymized data.

Quiz: Cloud Deployment Models for IoT

108.15 What’s Next?

Now that you understand deployment models, continue with:

Next Topic	Description
Cloud Security for IoT	IAM, encryption, and security best practices for cloud IoT
Cloud Platforms and Message Queues	Compare AWS, Azure, and messaging technologies