Serverless Functions Azure, AWS, GCP

Comparison of Function-as-a-Service offerings of AWS, Azure, and GCP

Transcript

1. Serverless Functions Azure, AWS, GCP Cloud Republic Event | Utrecht | October 30, 2019
2. • Software developer • Cloud • Serverless • Functional programming • Microsoft Azure MVP https://mikhail.io @MikhailShilkov Mikhail Shilkov
3. Function-as-a-Service Cloud Offerings
4. Hosting Plans Configurability Languages Programming Model Concurrency Isolation Cost Orchestrations Cold Starts Scalability Conclusions Q&A
5. Hosting Plans
6. Azure Functions
7. AWS Lambda and Google Functions
8. Configurability
9. Azure Functions Consumption
10. AWS Lambda Memory Allocation
11. GCF Memory Allocation
12. Azure Functions Premium
13. Programming Languages
14. Language Azure Functions AWS Lambda Google Cloud Functions .NET (C#/F#) GA GA - Node.js (JS/TS) GA GA GA JVM (Java) GA GA - Python GA (Linux only) GA GA PowerShell Preview GA - Go - GA GA Ruby - GA - Comparison Table
15. Programming Model
16. Azure Functions: Triggers and Bindings [FunctionName("MyFunc")] public static void Work( [TimerTrigger("0 */10 * * * *")] TimerInfo timer, [Blob("filename")] Stream stream, [Queue("myqueue")] out string message) { // body }
17. AWS: JSON Input and Output exports.handler = async (event) => { const response = { statusCode: 200, body: JSON.stringify('Hello from Lambda!'), }; return response; };
18. ASP.NET Core on AWS Lambda public class LambdaFunction : Amazon.Lambda.AspNetCoreServer.APIGatewayProxyFunction { protected override void Init(IWebHostBuilder builder) { builder .UseContentRoot(Directory.GetCurrentDirectory()) .UseStartup() .UseApiGateway(); } }
19. GCP: Express-like API exports.helloWorld = (req, res) => { let message = req.query.message || req.body.message || 'Hello World!'; res.status(200).send(message); };
20. Deployment Artifacts
21. Concurrency and Isolation
22. Resource Pool
23. Resource Pool
24. Resource Pool
25. Resource Pool
26. Resource Pool
27. Resource Pool
28. Resource Pool
29. Client Value Resource Utilization
30. Tradeoff
31. Server, VM, container, process, ...
32. Server, VM, container, process, ... VM VM
33. Server, VM, container, process, ... VM VM
34. Server, VM, container, process, ... VM VM
35. Server, VM, container, process, ... VM VM
36. Azure Functions 12
37. Legacy of App Service
38. Legacy of App Service
39. App Service: scalability
40. App Service: scalability
41. Function App Consumption Plan
42. Isolation Layers
43. Simultaneous Executions
44. Simultaneous Executions
45. Simultaneous Executions
46. Simultaneous Executions
47. AWS Lambda 22
48. AWS Lambda
49. AWS Lambda
50. AWS Lambda
51. Integrating with AWS Lambda
52. Simultaneous Executions in AWS Lambda
53. Isolation Layers
54. Simultaneous Executions in AWS Lambda
55. Simultaneous Executions in AWS Lambda
56. Simultaneous Executions in AWS Lambda
57. • • • • AWS Firecracker
58. Isolation Layers with Firecracker
59. «Naive» Function Composition
60. Statistical Multiplexing
61. Google Cloud Functions 31
62. Google Cloud Functions
63. Cost
64. Nominal Pricing
65. Nuances
66. Azure Simultaneous Executions
67. AWS Lambda: HTTP Integration
68. AWS Lambda: HTTP Integration
69. Orchestration
70. Azure Logic Apps
71. AWS Step Functions
72. Azure Durable Functions public static async Task Sequential(DurableOrchestrationContext context) { var conf = await context.CallActivityAsync<ConfTicket> ("BookConference", "ServerlessDays"); var flight = await context.CallActivityAsync<FlightTickets> ("BookFlight", conf.Dates); await context.CallActivityAsync("BookHotel", flight.Dates); }
73. Performance and Scalability
74. Ideal Scalability: Throughput
75. Ideal Scalability: Latency
76. Customer Value Cloud Resource Utilization
77. Cold Starts
78. Example: Loading a map
79. Example: Loading a map
80. Example: Loading a map
81. Cold Start
82. Warm Start
83. Period Before a Subsequent Cold Start
84. Cold Starts: Language Comparison
85. Dependencies Add to Cold Start
86. Fighting Cold Starts: Azure [FunctionName("Warmer")] public static void WarmUp( [TimerTrigger("0 */10 * * * *")] TimerInfo timer) { // No need to do anything }
87. AWS: Single instance warming
88. AWS: Multiple instance warming
89. ETL AT SCALE: Asynchronous data processing
90. Async Pipelines
91. Experiments with Queues
92. AWS Lambda Processing 100k SQS Messages
93. GCF Processing 100k Pub/Sub Messages
94. Azure Functions Processing 100k Queue Messages
95. CPU-intensive workload on AWS Lambda
96. CPU-intensive workload on Google Cloud Functions
97. CPU-intensive workload on Azure Functions
98. AWS Processing Speed per Reserved RAM
99. HTTP AT SCALE: Serving traffic of StackOverflow
100. Serving StackOverflow-like traffic
101. Requests/sec during the load test
102. 0 200 400 600 800 1000 1200 Azure Functions: Response Percentiles (ms)
103. 0 200 400 600 800 1000 1200 Azure Functions: Response Percentiles (ms)
104. 0 200 400 600 800 1000 1200 Azure Functions: Instances
105. Google Functions: Response Percentiles (ms)
106. Google Functions: Response Percentiles (ms)
107. 0 200 400 600 800 1000 1200 Google Functions: Instances
108. 0 200 400 600 800 1000 1200 AWS Lambda: Response Percentiles (ms)
109. 0 200 400 600 800 1000 1200 AWS Lambda: Response Percentiles (ms)
110. 0 200 400 600 800 1000 1200 AWS Lambda: Instances
111. So, What Should You Choose?
112. • • • • Follow-up reading (1)
113. • • • Follow-up reading (2)