Lessons learned while going serverless in production
0 likes679 views
- The document discusses lessons learned from going serverless in production including potential cost savings, hidden costs of serverless functions, and best practices for serverless applications. - It emphasizes starting simple and iterating, adopting an agile mindset, following AWS best practices, and sharing knowledge gained from experiences. - Key considerations for serverless applications include memory allocation impacts on costs, monitoring cloudwatch logs usage, and using containers and serverless functions together in architectures when applicable.
Lessons learned while going serverless in production
- 1. Lessons learned while going Serverless in production Mohamed Labouardy - Foxintelligence
- 2. Mohamed Labouardy Lead DevOps Engineer/Software Engineer Founder of Komiser - https://komiser.io Open Source contributor: Docker, Jenkins, Telegraf, etc DevSecOps Evangelist @mlabouardy https://labouardy.com
- 3. Detect Potential Cost Savings on Public Cloud +2K stars +185K downloads “Stay under budget by uncovering hidden costs, monitoring increases in spend, and making impactful changes based on custom recommendations.” https://komiser.io
- 4. How we do it ? Millions of real online transaction emails anonymized and processed in real time from a panel of 2.5M consumers DATA ACQUISITION & COLLECTION Billions of datapoints processed, cleaned and enriched through machine learning, scrapping, and blending DATA PROCESSING & ENRICHMENT Business acumen and analytical skills combined to provide game-changing insights to our clients on an intuitive platform DATA ANALYSIS & VISUALIZATION
- 6. Platform as Service - PaaS
- 10. CI/CD for Dockerized Microservices
- 12. Serverless Functions & Dockerized Microservices 10% Tiny part of our architecture
- 14. Foxintelligence Platform Statistics AWS Lambda AWS EKS API Gateway ElasticSearch +2B Invocations +5K Containers +80M Requests +5TB of Logs
- 18. Testing - ● SAM Local for testing locally ● Sandbox account for post-integrations tests ○ Best way to test is on the AWS infrastructure itself ● A/B testing with alias and weighted versions ● Performance testing for memory allocation ○ Apache benchmark ● Quality, unit & security tests included on CI/CD AWS Lambda Power Tuning
- 19. Deployment Framework ● CloudFormation extension optimized for Serverless ● Supports anything CloudFormation supports ● Infrastructure as Code
- 20. CI/CD for Lambda based Applications
- 22. CI/CD for Lambda based Applications
- 23. CI/CD for Lambda based Applications Jenkinsfile Dockerfile Audit dependencies to keep bundle sizes small
- 24. CI/CD for Lambda based Applications
- 26. Near Real Time Monitoring
- 27. Security ● Enable traceability ● Use most-restrictive permissions when setting IAM policies ● Protect secrets in transit and at rest ○ SSM Parameter Store, KMS ○ Vault ● Apply security at all layers ○ Code reviews, versioning ● Inject security and compliance into CI/CD ○ Dependency checks CVE/CVSE ○ Static/Dynamic analysis
- 28. Cold Start vs Warm Start ● Factors which increase the cold-start time: ○ Runtime environment ○ Memory size ○ Bundle size ○ VPC ● Solutions: ○ Programming Language ○ Warm-up ○ AWS Hyperplane
- 29. Cold Start vs Warm Start
- 30. Caching and Reuse Best Practices Connecting from AWS Lambda - MongoDB Atlas context.callbackWaitsForEmptyEventLoop=false
- 31. Serverless Hidden Costs “Memory allocation can make a big difference in your Lambda function cost. Too much allocated memory and you’ll overpay. Too little and your function will be at risk of failing. Therefore, you want to keep a healthy balance when it comes to memory allocation.”
- 32. Cloudwatch Logs eating up money
- 33. Key Takeaways ● Serverless or Containers? Why not both ● Be aware of Serverless hidden costs ● Start simple and iterate ● Adopt agile mindset ● Follow AWS best practices ● Share what you learn
- 34. @mlabouardy Follow me on Medium - @mlabouardy