Cut AWS Costs by 57% Using Spot Instances, Lambda@Edge, and CloudFront—Without Sacrificing Performance

AWS provides immense flexibility and scalability, but without proper optimisation, costs can quickly spiral out of control. Businesses often focus only on monetary costs (the direct AWS bill), but true cost optimisation requires considering?non-monetary?factors such as performance, reliability, and operational complexity.

This article explores the different types of costs in AWS and presents a?real-world cost breakdown comparison, demonstrating how strategic optimisations can reduce expenses by over?50%.

Types of AWS Costs

Monetary Costs (Direct AWS Bill)

These are the costs that appear on your AWS invoice. They can be broken down into:

• Compute costs?– EC2, Lambda, and containerised workloads.
• Storage costs?– S3, EBS, and Glacier.
• Data transfer costs?– Egress fees when moving data outside AWS.
• Networking costs?– VPC peering, transit gateways, and NAT gateways.
• Other service costs?– AI/ML processing, database usage, API Gateway, etc.

Non-Monetary Costs (Hidden & Indirect Costs)

These costs don’t show up directly on your bill but impact business efficiency, productivity, and customer experience:

• Operational complexity?– Managing infrastructure, Spot instance interruptions, and scaling challenges.
• Performance latency?– Higher latency results in a poor user experience.
• Reliability & availability?– A lower-cost architecture might have more failure points.
• Security & compliance?– Some cost-saving methods might introduce security trade-offs.
• Time-to-market?– Faster deployment leads to business agility and revenue growth.

To illustrate the impact of cost optimisation strategies, let’s consider a?video streaming platform?that needs:        

• Compute power for video encoding/transcoding.
• Edge computing for personalised content delivery.
• Content distribution for a global audience.

Assumptions

Compute Costs (On-Demand vs. Spot Instances)

On-Demand EC2 (Baseline Scenario)

• Instance type: c5.4xlarge (16 vCPUs, 32 GB RAM) for video encoding
• Hourly rate (On-Demand, US East 1): ~$0.68/hour
• Estimated usage: 24/7 operation (~730 hours/month)
• Monthly cost: 0.68 * 730 = $496.40 per instance
• Assumption: 2-3 instances required → $1,200 total

Spot Instance Pricing (Optimised Scenario)

• Spot pricing for c5.4xlarge: ~70-80% cheaper than On-Demand
• Estimated Spot cost: ~$0.18/hour
• Monthly cost: 0.18 * 730 = $131.40 per instance
• 2-3 instances required → ~$300 total (including capacity-optimised bidding overhead)
• Savings: ~75%

API Backend Compute (On-Demand vs. Lambda@Edge)

On-Demand EC2 Backend (Baseline)

• Instance type: t3.medium (2 vCPUs, 4 GB RAM)
• Hourly rate: ~$0.04/hour
• Monthly cost (autoscaled, 3 instances avg.): 0.04 730 3 = $87.60
• Estimated additional costs: Scaling + overhead → $300 total

Lambda@Edge (Optimised)

• Assumption: 1M requests per day (~30M per month)
• Request cost: $0.60 per million invocations
• Compute cost: $0.0000052 per 50ms of execution
• Estimated execution time: ~100ms per request
• Total cost: (30M 0.60/1M) + (30M 0.0000052 * 2) = $50
• Savings: ~83%

Content Delivery (Direct S3/EC2 vs. CloudFront)

Direct S3/EC2 Egress (Baseline)

• Data transfer out: 100 TB outbound
• Standard AWS egress pricing: $0.09 per GB
• Monthly cost: 100 TB 0.09 1024 = $9,216 (~$9,000 rounded)

CloudFront Egress (Optimised)

• CloudFront reduces origin traffic by ~50% (Cached Requests)
• CloudFront egress pricing: $0.02 - $0.085 per GB (Tiered)
• Assumed avg. cost: ~$0.04 per GB
• Monthly cost: 100 TB 0.04 1024 = $4,096 (~$4,000 rounded)
• Savings: ~56%

CloudFront vs. Direct S3 Serving

Baseline: Serving Video Content via EC2/S3

• EC2 API requests + S3 GET requests: $1,000 (Estimate based on large-scale traffic)

Optimised: CloudFront + S3 Origin

• CloudFront caching reduces S3 GET requests → Lowers costs
• CloudFront request pricing: ~$0.0075 per 10,000 requests
• Estimated monthly cost: ~$600
• Savings: ~40%

The cost estimates were derived using AWS public pricing, capacity-optimised bidding models, and industry best practices for scaling workloads efficiently. These numbers are approximate but based on real-world usage patterns. If you want an even more precise estimate for your specific workload, we could use AWS Cost Calculator or analyse a live AWS bill.

Cost Breakdown: Standard AWS Architecture vs. Optimised AWS Architecture

Baseline Cost Scenario (Without Optimisation)

Optimised Cost Scenario (Using Spot Instances, Lambda@Edge, and CloudFront)

Key Learnings and Business Impact

Compute Savings with Spot Instances

• Using Spot Instances for video encoding resulted in 75% savings.
• The capacity-optimised bidding strategy ensured that workloads had minimal interruptions.
• Trade-off: Requires automation to handle interruptions.

Serverless Edge Computing with Lambda@Edge

• Lambda@Edge eliminated the need for EC2 backend servers for API requests, cutting backend compute costs by 83%.
• Non-monetary benefit: Lower latency for global users.
• Trade-off: Occasional cold start delays.

Content Delivery Optimisation with CloudFront

• Using CloudFront as a CDN reduced S3/EC2 data transfer costs by 56%.
• Non-monetary benefit: Faster load times and improved reliability.
• Trade-off: Some caching limitations for dynamic content.

Final Thoughts: Why This Matters

By combining?Spot Instances, Lambda@Edge, and CloudFront, we reduced AWS costs from?$11,500 to $4,950?per month, a?57% savings—all while improving performance and scalability.

Takeaway:?AWS cost optimisation isn’t just about reducing your bill—it’s about aligning costs with business goals while maintaining high performance, security, and availability.

???What AWS cost-saving strategies have worked for your business? Let’s discuss in the comments!