Saving Money on Serverless: Common Mistakes and How to Avoid Them

Join us in this insightful webinar as we delve into the world of serverless architecture, exploring common pitfalls that often lead to unnecessary costs. Yan Cui, AWS Serverless Hero, and Alex de Boutray, VP of Engineering at Theodo, will navigate through key strategies and best practices to optimize your serverless infrastructure.

This session is a must-attend for cloud enthusiasts, developers, and decision-makers seeking actionable insights to maximize cost efficiency in their serverless environments. Gain practical knowledge, learn from real-world examples, and equip yourself with the tools to avoid common pitfalls and save money on your serverless journey.

Transcript

2. “serverless”
3. “serverless”
4. www.cloudzero.com/blog/our-take-on-the-new-stack-survey
5. serverless is
6. serverless is SCALABLE
7. serverless is RESILIENT
8. serverless is SECURE
9. COST-EFFICIENT serverless is
10. AWS is hard.
11. Mistakes can be expensive.
14. Trace through DynamoDB streams, SNS, SQS, Kinesis, EventBridge, etc.
15. RDS, Redis, Mongodb, Postgres, MS SQL, …
16. RDS, Redis, Mongodb, Postgres, MS SQL, …
18. BEST Serverless Observability Tool!!
19. Yan Cui AWS Serverless Hero Independent Consultant Developer Advocate @ Lumigo Alex de Boutray VP of Engineering @ Theodo
21. overlooking monitoring and logging costs
22. CloudWatch often cost much more than your core application.
23. As cost goes up, value goes down.
24. How to keep CloudWatch Logs cost under control 1. Do structured logging.
25. How to keep CloudWatch Logs cost under control 1. Do structured logging. DEBUG INFO WARN ERROR Detailed events for debugging application. General information that highlights progress of application. Potential problems, but doesn’t stop application from working. Issues that require immediate attention.
26. How to keep CloudWatch Logs cost under control 1. Do structured logging. Log at INFO or above in production. DEBUG INFO WARN ERROR Detailed events for debugging application. General information that highlights progress of application. Potential problems, but doesn’t stop application from working. Issues that require immediate attention.
27. How to keep CloudWatch Logs cost under control 1. Do structured logging. Log at INFO or above in production. 2. Sample DEBUG logs in production. e.g. 5% of invocations.
28. How to keep CloudWatch Logs cost under control 1. Do structured logging. Log at INFO or above in production. 2. Sample DEBUG logs in production. e.g. 5% of invocations. 3. Set log retention to 30 days.
30. CloudWatch Logs AWS Lambda AWS Lambda stdout asynchronously invokes any log aggregation service
32. CloudWatch Logs AWS Lambda AWS Lambda stdout asynchronously invokes any log aggregation service double paying for logs
33. Lambda Extensions + Telemetry API
34. https://docs.aws.amazon.com/lambda/latest/dg/telemetry-api.html
36. CloudWatch Logs AWS Lambda
37. Log aggregation is coming to Lumigo soon.
43. DEBUG INFO WARN platform.initStart platform.start platform.initRuntimeDone platform.initReport platform.runtimeDone platform.report unhandled exception
44. DEBUG INFO WARN platform.initStart platform.start platform.initRuntimeDone platform.initReport platform.runtimeDone platform.report unhandled exception
45. not having billing alarms
46. Everyone should use billing alarms.
47. They are not perfect. But they can still save your a$$.
51. not reading documentations
52. Every AWS service has a pricing page. Read it!
60. 40M DynamoDB reads = ~$10
61. over-provisioning
62. More memory = more CPU = more network bandwidth
64. Easy to be wrong by an order of magnitude.
65. STORY TIME
68. Right-sizing Lambda functions
69. https://github.com/alexcasalboni/aws-lambda-power-tuning
70. Use ARM architecture
71. Use ARM architecture (after testing…)
72. 25% cheaper
73. Best for functions with a lot IO wait time.
74. missing caching
75. Caching is a cheat code for building performant & scalable applications.
76. bit.ly/sls-caching
77. aaronstuyvenberg.com/posts/containers-on-lambda
78. www.usenix.org/system/ fi les/atc23-brooker.pdf
79. bit.ly/social-network-in-4-weeks
80. average 99% cache hit rate
82. synchronous lambda-to-lambda invocations
87. SYNCHRONOUS Lambda-to-Lambda are almost always a sign of bad design.
89. Service Boundary
90. Service Boundary
91. Service Boundary
92. AWS Lambda function !== lambda function in programming
94. ?
95. ? ?
97. poor architecture design
98. Every architectural decision is a buying decision.
99. Using the wrong service can be very costly.
102. www.thefrugalarchitect.com
103. Law I. Make Cost a Non-functional Requirement. Law II. Systems that Last Align Cost to Business. Law III. Architecting is a Series of Trade-offs. Law IV. Unobserved Systems Lead to Unknown Costs. Law V. Cost Aware Architectures Implement Cost Controls. Law VI. Cost Optimization is Incremental. Law VII. Unchallenged Success Leads to Assumptions.
104. Law I. Make Cost a Non-functional Requirement. Law II. Systems that Last Align Cost to Business. Law III. Architecting is a Series of Trade-offs. Law IV. Unobserved Systems Lead to Unknown Costs. Law V. Cost Aware Architectures Implement Cost Controls. Law VI. Cost Optimization is Incremental. Law VII. Unchallenged Success Leads to Assumptions.
105. IOT core AppSync API Gateway Messages Connection Time $2 per million $0.08 per million mins $0.25 per million mins $1 per million $0.08 per million mins $1 per million
106. Users Connection Time
107. Users Connection Time Revenue
108. Users Connection Time Revenue Engagement
109. Users Connection Time Revenue Engagement
110. www.gomomento.com/services/topics
112. IOT core AppSync API Gateway Messages Connection Time $2 per million $0.08 per million mins $0.25 per million mins $1 per million $0.08 per million mins $1 per million Momento - - Data Transfer $0.5 / GB EC2 rates EC2 rates EC2 rates
113. Adding necessary moving parts to your architecture.
114. ? ?
115. ? ?
116. Avoid unnecessary moving parts to your architecture.
117. The most dangerous phrase in the language is "we've always done it this way". - Grace Hopper
118. 1. Overlooking monitoring and logging costs 2. Not having billing alarms 3. Not reading documentations 4. Over-provisioning 5. Missing caching 6. Synchronous lambda-to-lambda calls 7. Poor architecture design
119. Get Started Troubleshooting Your Serverless Databases Today - Free Lumigo.io/signup
120. Questions?