AWS Pricing & Cost Management

Imagine AWS is like a massive hotel that also provides electricity (compute) and cupboards (storage).

• On-Demand: You walk into the hotel without a booking. You stay for one night, pay for one night, and leave. It’s flexible but costs the standard rate.
• Reserved Instances: You tell the hotel manager, “Sir, I will definitely stay here for 1 year.” Because you promised (committed), they give you a huge discount (up to 75% off).
• Spot Instances: The hotel has empty rooms at night that are going to waste. They offer them to you for a very cheap price (up to 90% off), but the catch is, if a VIP guest (On-Demand user) comes, you have to vacate the room in 2 minutes.

In simple terms: AWS charges you mainly for three things: running servers (Compute), saving files (Storage), and sending data out to the internet (Outbound Data Transfer).

AWS pricing might look complicated, but it basically works on a “Pay-as-you-go” model. This means you don’t need to buy expensive servers upfront. You just rent them. If you switch off the server, the billing stops (mostly). It is just like your electricity bill at home you pay for what you consume.

The 3 Main Cost Drivers:

1. Compute: This is the cost for the “brain” power. If your website is running, the meter is running.
2. Storage: This is the cost for the “hard drive” space. Even if nobody visits your website, you pay to keep your files stored there.
3. Outbound Data Transfer: Bringing data into AWS is free. Moving data between AWS services (in the same region) is usually free. But sending data out to the internet (like when someone downloads a file from your site) costs money.

Tools you should know:

• To check prices before using: Use the AWS Pricing Calculator. It helps you estimate your monthly bill before you even start.
• To see the bill: Use AWS Cost Explorer. It shows graphs of where your money is going.
• To get alerts: Use AWS Budgets. You can set a limit (e.g., $10), and if you cross it, AWS sends you an email.

DevSecOps Architect Level

Concept: As an architect, your goal isn’t just “cheap” it is Cost Optimization. You need to balance performance with price. You must leverage “Commitment Pricing” (Savings Plans) and “Spot Instances” for stateless workloads to reduce OpEx (Operating Expense).

Deep Dive into Pricing Models:

1. On-Demand: Good for “spiky” workloads or new apps where you don’t know the traffic patterns yet. No commitment.
2. Savings Plans (The Modern Standard):
   • Compute Savings Plans: Most flexible. You commit to spending, say $5/hour. Applies to EC2, Fargate, and Lambda across any Region. Recommended for microservices architectures.
   • EC2 Instance Savings Plans: Less flexible, deeper discount. You commit to a specific instance family (e.g., M5) in a specific Region (e.g., Mumbai).
3. Reserved Instances (RIs): Older model but still relevant for RDS (Databases), Redshift, and ElastiCache. You buy a “coupon” for capacity.
   • Strategy: Use Partial Upfront for a balance of cash flow and discount.
4. Spot Instances: Use these for CI/CD pipelines, batch processing, or containerized stateless apps (EKS). Handle interruptions using “Spot Instance Termination Notices” (2-minute warning).

• AWS Trusted Advisor: Scans your infrastructure and tells you if you have “Zombie resources” (e.g., unattached EBS volumes or Idle Load Balancers) that are wasting money.
  • Official Trusted Advisor
• AWS Cost & Usage Report (CUR): The most granular data available. You can feed this into Amazon Athena to run SQL queries on your billing data.
  • Official AWS CUR Documentation

—

Use Case

Scenario: A Startup launching a Video Streaming App in India.

• Phase 1 (Development): The team uses On-Demand instances because they are testing and turning servers off frequently. They use AWS Free Tier for the first 12 months.
• Phase 2 (Production – Steady State): The app has a steady user base. The Architect buys a Compute Savings Plan for the base load (the minimum servers always running) to save roughly 66%.
• Phase 3 (Viral Event): A big cricket match streams. Traffic spikes. The Auto Scaling Group launches Spot Instances to handle the extra viewers cheaply, as the video transcoding job is fault-tolerant.

Benefits

1. No Upfront Capital: Startups don’t need crores of rupees to buy data centers.
2. Agility: You can experiment. If a project fails, shut it down. You only pay for what you used.
3. Scale: You can go from 1 server to 1,000 servers in minutes during high traffic.

—

Technical Challenges

• Bill Shock: The most common issue. Beginners often forget to turn off an expensive instance (like a large EC2 or RDS) and wake up to a huge bill.
• Zombie Resources: Deleting an EC2 instance does not always delete the attached storage (EBS volume) or Elastic IP. These keep charging money silently.
• Data Transfer Costs: If your architecture spans multiple regions (e.g., database in US, app in India), the data transfer cost between regions can become very high unexpectedly.

• AWS Pricing Overview: https://aws.amazon.com/pricing/
• AWS Free Tier Details: https://aws.amazon.com/free/
• Cost Management Service: https://aws.amazon.com/aws-cost-management/

8. Cheat Sheet (Easy Remember Table)