While preparing for the AWS SAA-C03, many candidates get confused by data migration strategies. In the real world, this is fundamentally a decision about Network Bandwidth Economics vs. Physical Transport Logistics. Let’s drill into a simulated scenario.
The Scenario #
MediaVault Productions operates a regional media processing center with 70 TB of archived video content stored on Network-Attached Storage (NAS) using NFS protocol. File sizes range from 1 MB (promotional clips) to 500 GB (raw 4K footage). The archive has reached its final size—no new content will be added. Leadership has mandated migration to cloud object storage to decommission the on-premises infrastructure, with two critical constraints: minimize migration duration and avoid saturating the existing 1 Gbps internet connection that supports active production workflows.
Key Requirements #
Migrate 70 TB of NFS-based video files to Amazon S3 with minimal network bandwidth consumption and fastest possible completion time.
The Options #
- A) Create an S3 bucket with appropriate IAM permissions and use AWS CLI to copy all files directly over the internet connection.
- B) Order an AWS Snowball Edge device, transfer data locally to the appliance, then ship the device back to AWS for S3 import.
- C) Deploy an AWS Storage Gateway (File Gateway) on-premises with a public service endpoint, create an NFS share backed by S3, and copy data through the gateway over the internet.
- D) Provision AWS Direct Connect between the data center and AWS, deploy Storage Gateway with a public Virtual Interface (VIF), create an NFS share backed by S3, and transfer data through the gateway.
Correct Answer #
Option B.
Step-by-Step Winning Logic #
This scenario contains two hard constraints that eliminate network-based solutions:
- “Minimize network bandwidth” — This explicitly rules out any solution that transfers 70 TB over the internet (Options A, C).
- “Complete as quickly as possible” — Direct Connect (Option D) requires 2-4 weeks for circuit provisioning alone, plus ongoing costs.
Snowball Edge is purpose-built for this exact use case:
- 80 TB usable capacity per device (sufficient for the entire dataset)
- Local transfer speed: 10 Gbps network interface allows ~1-2 days for data copy
- Total migration time: 5-10 days including shipping logistics
- Zero network impact: Completely offline from production bandwidth
- Cost-effective for one-time migration: ~$300-400 flat fee vs. weeks of Direct Connect charges
The Service Selection Principle (Associate Level Focus): When you see “large dataset” + “minimize bandwidth” + “one-time migration”, AWS Snowball family is the default answer. For datasets >10 TB, physical transport beats network transfer in both speed and cost.
💎 The Architect’s Deep Dive: Why Options Fail #
Correct Answer #
Option B — AWS Snowball Edge for physical data transport.
The Traps (Distractor Analysis) #
-
Why not Option A (AWS CLI Direct Transfer)?
- Bandwidth Violation: Transferring 70 TB over 1 Gbps saturates the circuit for 12-18 days, directly violating the “minimize bandwidth” requirement.
- Hidden Cost: While free for data ingress, this monopolizes business-critical connectivity.
- Exam Trap: Tests if you recognize when “simple” solutions violate constraints.
-
Why not Option C (Storage Gateway over Internet)?
- Same bandwidth problem: Storage Gateway still transfers 70 TB over the internet—it just adds an abstraction layer with no benefit.
- Unnecessary complexity: Adds gateway appliance management for a one-time migration.
- Cost inefficiency: Gateway EC2 instance + bandwidth saturation vs. flat Snowball fee.
-
Why not Option D (Direct Connect + Storage Gateway)?
- Time violation: Direct Connect provisioning takes 2-4 weeks minimum (often 6+ weeks in practice), failing the “as quickly as possible” requirement.
- Cost overkill: $0.30/hour port fee + $0.02/GB data transfer for a one-time 70 TB job = $1,400+ in transfer costs alone, plus ongoing monthly port fees.
- Over-engineering: Direct Connect makes sense for ongoing hybrid workflows, not one-time migrations.
- Exam Lesson: Associate-level questions penalize solutions that introduce recurring costs for temporary needs.
The Architect Blueprint #
graph TD
A[On-Premises NAS
70 TB NFS Storage] -->|1. Order Device| B[AWS Snowball Edge
80 TB Capacity]
B -->|2. Ship to Data Center
2-3 days| C[Local Data Transfer
10 Gbps Interface]
A -->|3. Copy 70 TB
1-2 days| C
C -->|4. Ship to AWS
2-3 days| D[AWS Import Facility]
D -->|5. Data Load to S3
1-2 days| E[Amazon S3 Bucket
Video Archive]
style B fill:#FF9900,stroke:#232F3E,stroke-width:3px,color:#fff
style E fill:#569A31,stroke:#232F3E,stroke-width:3px,color:#fff
style C fill:#FF9900,stroke:#232F3E,stroke-width:2px
Diagram Note: Snowball Edge enables a 5-10 day total migration timeline with zero impact on production network bandwidth—the data never traverses the internet connection.
The Decision Matrix #
| Option | Est. Complexity | Est. Total Cost | Time to Complete | Network Impact | Pros | Cons |
|---|---|---|---|---|---|---|
| A (AWS CLI) | Low | $0 (ingress free) | 12-18 days | CRITICAL (100% saturation) | Zero service cost; Simple tooling | Violates bandwidth constraint; Blocks production traffic |
| B (Snowball Edge) ✅ | Medium | ~$350-400 | 5-10 days | None | Fastest; Zero bandwidth use; Fixed cost | Requires physical handling; 2-3 day shipping each way |
| C (Storage Gateway + Internet) | High | ~$150 (Gateway) + bandwidth saturation | 12-18 days | CRITICAL (100% saturation) | NFS interface compatibility | Adds complexity with no bandwidth benefit; Violates constraint |
| D (Direct Connect + Gateway) | Very High | ~$2,000+ (DX setup + transfer + monthly fees) | 21-45 days (provisioning) | Low (dedicated circuit) | High throughput; Secure; Reusable | Extreme cost for one-time use; Fails speed requirement; Ongoing fees |
Cost Breakdown (Option B - Snowball Edge):
- Device fee: ~$300 per job (first 10 days included)
- Shipping: Included in device fee
- Data transfer OUT of Snowball into S3: $0 (free)
- Total: ~$300-400 one-time cost
Cost Breakdown (Option D - Direct Connect):
- Port fee: $0.30/hour × 720 hours = ~$216/month
- Data transfer: 70,000 GB × $0.02/GB = $1,400
- Provisioning time cost: 3-6 weeks of delayed infrastructure decommissioning
- Total first month: ~$1,600+ (plus ongoing $216/month if not terminated)
Real-World Practitioner Insight #
Exam Rule #
“For the SAA-C03 exam, when you see ’large dataset migration’ (>10 TB) + ‘minimize bandwidth’ + ‘one-time move’, immediately eliminate network-based options and select Snowball family devices. If the dataset is >80 TB, choose Snowball Edge with multiple devices or consider Snowmobile (100 PB scale).”
Real World #
In production environments, I’ve seen organizations make these adjustments:
-
Hybrid Approach: Use Snowball for the bulk 70 TB migration, then set up Storage Gateway for the ongoing 2-5 GB/day of new content that starts accumulating after “final” archive claims.
-
DataSync Consideration: For datasets under 10 TB or when you do have unused bandwidth, AWS DataSync offers better automation than raw CLI commands—it handles retries, verification, and metadata preservation.
-
Multi-Device Parallelization: For time-critical migrations >80 TB, order multiple Snowball devices simultaneously and partition data to achieve 2-3 day migration windows.
-
Direct Connect Reality Check: We only provision Direct Connect when there’s a multi-year hybrid cloud strategy. The 45-day provisioning time cited in Option D is optimistic—I’ve seen 90+ days for complex network approvals.
-
Cost Psychology: The “$0” ingress cost of Option A psychologically appeals to finance teams until you quantify the opportunity cost of 12-18 days of degraded business operations—often worth $10,000+ in lost productivity.
The Associate-Level Lesson: Focus on matching service purpose to requirement. Snowball exists specifically to solve the “big data, small pipe” problem. Don’t overthink it.