"It’s not only Lambda! Economics behind Serverless" at Serverless Architecture Conference April 2019 by Christian Bannes and Vadym Kazulkin (both ip.labs GmbH)
- 1. It‘s not only Lamda Economics behind serverless by Christian Bannes and Vadym Kazulkin, ip.labs GmbH
- 2. Contact Vadym Kazulkin ip.labs GmbH v.kazulkin@iplabs.de xing.com/profile/Vadym_Kazulkin @VKazulkin @ServerlessBonn (Meetup) Christian Bannes ip.labs GmbH c.bannes@iplabs.de xing.com/profile/Christian_Bannes
- 3. Ip.labs GmbH
- 4. Money for nothing Developers are good guys They can ignore money... Image: https://www.flickr.com/photos/pictures-of-money/17121925920/
- 7. It‘s not only Lamda… ...also consider the infrastructure Image: https://stocksnap.io
- 8. You also need… • Storage • Databases • Queues • Api Gateway • Logging • Data Transfer • And More ... Image: https://stocksnap.io
- 9. Most expensive… ...are people Image: https://burst.shopify.com
- 10. What you really have to consider: Total Cost of Ownership
- 12. Cost for lambda REQUEST DURATION
- 13. Request Tier $ 0.20 Per 1 Mio Requests
- 14. Duration Tier $ 0.00001667 Per GB-Second
- 15. GB-Second ONE SECOND ONE GB
- 16. Example • 1 Mio requests • Lambda with 512MB • Each lambda takes 200ms • 0.5GB * 0.2sec * 1Mio = 100 000 GB-Seconds Requests: $0.20 GB-Seconds: $1.67
- 18. How does this compare to VM pricing model?
- 19. Lambda performance compares to • M4 (general purpose) • M5 (general purpose) EC2 Instance Families
- 20. M5 General Purpose Instance Cost scales linearly with memory
- 22. Lambda compute is 5x more expensive than M5
- 23. Threading Model Idle Idle Pay per wall clock time Pay per use
- 24. AWS monetizes the datacenter… … by selling time on their servers Image: https://aws.amazon.com
- 25. Common pitfalls when working with lambda
- 26. More memory = more expensive?
- 27. Power tuning lambda • Executes different settings in parallel • Outputs the optimal setting Image: https://github.com/alexcasalboni/aws-lambda-power-tuning
- 28. Paying for wait time Image: https://blog.binaris.com/lambda-pricing-pitfalls/
- 29. The reality is… Lambda is often just a small percentage of your total cost
- 30. $ 3.50 Per million API calls Api Gateway
- 31. Event Sources • SQS and SNS are charged for requests • Kinesis charges for shard hours & PUT requests Image: https://blog.binaris.com/lambda-pricing-pitfalls/
- 32. Event Sources • Cost for Kinesis grows with slower rate • Attractive at to operate at scale Image: https://blog.binaris.com/lambda-pricing-pitfalls/
- 34. Provisioned vs On-Demand • Use On-Demand for spiky workloads • Use Provisioned for constantly high workload
- 35. Cost saving with On-Demand Image: https://twitter.com Switch to "On-Demand"
- 36. Example $2400 / Month $520 / Month 100 API requests per second 24/7 • API Gateway • Lambda with 1GB and 300ms per request • Load balancer • 3 x M5.xlarge with 16GB each
- 37. Serverless vs VM • "Pay as you go" can significantly reduce server cost because you don’t pay for ide time • Cost for compute is always higher • For serverless you need right workload and architecture • For constantly high workloads it can quickly become expensive
- 38. What you really have to consider: Total Cost of Ownership
- 40. TCO Full Picture Forrest Brazeal „The Business Case For Serverless” https://www.trek10.com/blog/business-case-for-serverless/ No Infrastructure Maintanence Auto Scaling Built in Fewer Engineers Required Lower Technical Debt Focus on Business Value & Innovation Faster Time to Market
- 41. Do more with (Server-)less
- 42. FaaS... …or not to FaaS
- 44. 1. Application lifecycle 2. Workloads 3. Platform limitations 4. Cost at scale 5. Operational constraints 6. Organizational knowledge Decision Checklist
- 45. Understand Application lifecycle Christian Posta „Be as serverless as you can, but not more than that” https://blog.christianposta.com/serverless/its-not-about-microservices-vs-serverless/
- 46. Explore phase • Quickly validate hypotheses • Rapidly experiment • Run experiments as cheaply as possible Image: burst.shopify.com/photos/a-look-across-the-landscape-with-view-of-the-sea Serverless is a perfect fit
- 47. Exploit phase • Built something that does provide customer value • Build it on scale • Build a profitable product around it Image: Robert Scoble via Flickr partly serverless and partly not serverless architecture
- 48. Application lifecycle • How much of my stack should I own to be able to deliver business value? • Outsource SLA, regulatory compliance, price, and roadmap to my service provider?
- 49. Existing applications • You can’t magically move that all off to service providers • You can try to modernize parts of them
- 50. FinDev Concept Activity-based costing on a digital operation-by-operation basis • Figure out features which deliver business value comparing to their cost Aleksander Simovic & Mark Schwarz „FinDev and Serverless Microeconomics: Part 1” https://aws.amazon.com/de/blogs/enterprise-strategy/findev-and-serverless-microeconomics-part-1/
- 51. 1. Application lifecycle 2. Workloads 3. Platform limitations 4. Cost at scale 5. Operational constraints imposed by other dependencies 6. Organizational knowledge
- 52. Understand your workloads • Event-driven/reactive • API-driven • Batch Job • ML/AI • Big Data Image: flickr.com/photos/everywhereatonce/294789504
- 53. Understand your workloads • Do we need to access specialized hardware ? • GPU access required? • Do we need constantly high performance? • Response time below 100 ms (e.g. bidding or gaming platforms) “A Berkeley View on Serverless Computing” https://www2.eecs.berkeley.edu/Pubs/TechRpts/2019/EECS-2019-3.html
- 54. Understand your workloads • Do we need high throughput ? • Lambda‘s network bandwidth is limited shared between all functions packed on the same VM • Do functions need to communicate with each other? • functions not directly network accessible, they must communicate via an intermediary service “A Berkeley View on Serverless Computing” https://www2.eecs.berkeley.edu/Pubs/TechRpts/2019/EECS-2019-3.html
- 55. 1. Application lifecycle 2. Workloads 3. Platform limitations 4. Cost at scale 5. Operational constraints imposed by other dependencies 6. Organizational knowledge
- 56. Understand platform limitations • Cold start • Lambda with and without VPC for each runtime • Invocation duration/ Timeouts • Lambda 15min • API Gateway integration 29sec • Max Memory • Lambda 3GB
- 57. Ajay Nair „Become a Serverless Black Belt” https://www.youtube.com/watch?v=oQFORsso2go
- 58. Ajay Nair „Become a Serverless Black Belt” https://www.youtube.com/watch?v=oQFORsso2go
- 59. Understand platform limitations • Cold start • Lambda with and without VPC for each runtime • Invocation duration/timeouts • Lambda 15min • API Gateway integration 29sec • Max Memory • Lambda 3GB
- 60. Understand platform limitations • Max concurrent invocations • Lambda 3000 per account • Scaling thresholds • Dynamo DB provisioned and on- demand capacities
- 61. 1. Application lifecycle 2. Workloads 3. Platform limitations 4. Cost at scale 5. Operational constraints imposed by other dependencies 6. Organizational knowledge
- 62. Understand your cost at scale • Lambda • API Gateway • Dynamo DB capacity choices • Remote API calls • Logging costs • Monitoring costs
- 63. 1. Application lifecycle 2. Workloads 3. Platform limitations 4. Cost at scale 5. Operational constraints imposed by other dependencies 6. Organizational knowledge
- 64. Understand the operational constraints imposed by other dependencies • Language runtime choice • Can be forced by some dependencies to other libraries • Relation database choice • Aurora Serverless: either use VPC or use Data API Yan Cui „Not so FaaS” https://blog.binaris.com/not-so-faas/
- 65. 1. Application lifecycle 2. Workloads 3. Platform limitations 4. Cost at scale 5. Operational constraints imposed by other dependencies 6. Organizational knowledge
- 66. Understand organizational knowledge • Do you already have DevOps knowledge? • Are the developers willing to learn new languages?
- 69. Berkeley View on Serverless Computing “A Berkeley View on Serverless Computing” https://www2.eecs.berkeley.edu/Pubs/TechRpts/2019/EECS-2019-3.html
- 70. Berkeley View on Serverless Computing • Provide low latency and high IOPS Serverless Ephemeral Storage • Provide Serverless Durable Storage • Improve Networking • Improve security • Accommodate cost-performance “A Berkeley View on Serverless Computing” https://www2.eecs.berkeley.edu/Pubs/TechRpts/2019/EECS-2019-3.html
- 71. Predictions • More available runtimes (e.g. GraalVM) • Platform limitations will be weakened or disappear • New use cases of serverless (Big Data, ML/AI) • Access to specialized hardware • Companies gaining advantages from the FinDev approach Image: Marco Verch via Flickr (Creative Commons 2.0) https://www.flickr.com/photos/30478819@N08/26528539689/
- 72. Serverless vs VM • "Pay as you go" can significantly reduce server cost because you don’t pay for ide time • Cost for compute is always higher • For serverless you need right workload and architecture • For constantly high workloads it can quickly become expensive • Application lifecycle • Workloads • Platform limitations • Cost at scale • Operational constraints • Organizational knowledge FaaS or not to FaaS
- 74. Thank You!