Most Important Cloud Concepts - Episode 4 - Serverless

Serverless: Evolution and Current Interpretation

Key Concepts:

• Serverless computing
• AWS Lambda
• EC2 instances
• Lambda functions
• Infrastructure management
• Pay-per-use model
• AWS OpenSearch
• Master nodes, Data nodes (OpenSearch terminology)
• Scaling
• Cost model

Original Definition of Serverless (Driven by AWS Lambda)

• Problem: Traditionally, deploying code (e.g., for web apps or scripts) required provisioning and managing EC2 instances (virtual machines). This involved setup, configuration, security maintenance, and networking, which was complex and expensive.
• AWS Lambda's Solution: Lambda functions abstract away the underlying infrastructure. Developers write code, upload it to a Lambda function, and the Lambda service handles execution without the developer needing to manage EC2 instances.
• Abstraction: Developers don't need to know how many EC2 instances are running their code (could be one, multiple, or none when idle). The Lambda service distributes traffic and scales instances automatically.
• Focus on Code: The primary goal is to allow developers to focus on writing code and business logic, not infrastructure management.
• Key Feature: No management of the underlying infrastructure.

Shift in the Meaning of Serverless (Example: AWS OpenSearch)

• Traditional OpenSearch: Previously, users had to configure OpenSearch clusters by specifying the number and type of instances (e.g., master nodes, data nodes).
• "Serverless" OpenSearch: AWS introduced a version of OpenSearch that automatically scales based on traffic. Users don't need to configure the underlying instances.
• The Catch: Unlike AWS Lambda's pay-per-use model, "serverless" OpenSearch still charges based on the underlying instances provisioned, even though the user doesn't manage them directly.
• Example: If the OpenSearch cluster scales to two instances costing $100 each, the user pays $200 per month, prorated. If it scales to six instances, the cost becomes $600 per month.
• Contradiction: This model contradicts the original serverless concept, where users only pay for actual usage (e.g., the number of function executions).

Arguments and Perspectives

• Original Serverless Ideal: True serverless (like AWS Lambda) means no infrastructure management and pay-per-use pricing.
• Current AWS Marketing: AWS increasingly uses "serverless" to describe services where infrastructure is managed by AWS, but users still pay for the underlying resources.
• Author's View: The author dislikes the expanded definition of serverless, finding it confusing and misleading.
• Call to Action: Be cautious when encountering the term "serverless," as its meaning can vary depending on the context and the speaker.

Conclusion

The term "serverless" has evolved. Originally, it meant a complete abstraction of infrastructure management and a pay-per-use pricing model (exemplified by AWS Lambda). However, AWS now uses the term more broadly to describe services where they manage the infrastructure, but users still pay for the underlying resources, regardless of actual usage (as seen in the "serverless" OpenSearch example). Therefore, it's crucial to understand the specific context when encountering the term "serverless" to avoid confusion and make informed decisions.