Alex Karcher 5 tips for production ready Azure Functions Program manager – Microsoft 5 tips for production ready Azure Functions

Common Architecture Patterns Serverless APIs Mobile / Web backends Event and Stream Processing IoT event and data processing Big data and machine learning pipelines Integration and Enterprise Service Bus Connecting line-of-business systems Pub/Sub for business events Automation and Digital Transformation Process automation

Tip 1: Serverless APIs and HTTP HaveiBeenPwnd, Quest, Starbucks

Sample synchronous serverless architecture Cosmos DB API Management (primary) Traffic Manager API Management (secondary) Cosmos DB

Azure Functions “Behind the scenes” HTTP Request HTTP Response

Functions “Behind the scenes” Azure Infrastructure Functions Instance Azure Functions Runtime HTTP Request Azure Functions (App Services) HTTP Front-end Functions Instance HTTP Response Azure Functions Runtime Scale Controller Functions Instance Azure Functions Runtime

Scalability Considerations The “Scale Controller” determines how many instances are actively doing work For Event Hub, Queue, Service Bus the scale controller can assess message backlog length and burndown rate to scale For HTTP functions today looks at request queue for each instance The “host.json” config file allows you to set instance limits for concurrency at an instance level Other knobs are to turn off file-system logging, and not include a dashboard storage account

Your App in Concept

Your app with long cold start

Your app with one pre-warmed instance

Scale out settings

Your app with one pre-warmed instance Minimum Plan Size

Your app with a 3 instance min plan size Minimum Plan Size

Tip 2: Event Stream Processing Starbucks, Honeywell, Walmart

Sample synchronous serverless architecture Cosmos DB Devices and Clients Event / IoT Hubs

Event Stream Processing: Scaling Starbucks, Honeywell, Walmart

Sample synchronous serverless architecture Devices and Clients Event / IoT Hubs Function Instances Cosmos DB

Sample synchronous serverless architecture Devices and Clients Event / IoT Hubs Function Instances Device 1 Partition 1 Partition 2 Cosmos DB Partition 3 Partition 4

Sample synchronous serverless architecture Devices and Clients Event / IoT Hubs Function Instances Device 1 Partition 1 Partition 2 Cosmos DB Partition 3 Partition 4

Sample synchronous serverless architecture Devices and Clients Event / IoT Hubs Function Instances Device 1 Partition 1 Device 2 Partition 2 Cosmos DB Device 3 Partition 3 Device 4 Partition 4

Sample synchronous serverless architecture Devices and Clients Event / IoT Hubs Function Instances Device 1 Partition 1 Device 2 Partition 2 Cosmos DB Device 3 Partition 3 Device 4 Partition 4 Device 5

Sample synchronous serverless architecture Devices and Clients Event / IoT Hubs Function Instances Device 1 Partition 1 Device 2 Partition 2 Cosmos DB Device 3 Partition 3 Device 4 Partition 4 Device 5

Sample synchronous serverless architecture Devices and Clients Event / IoT Hubs Function Instances Device 1 Partition 1 Device 2 Partition 2 Cosmos DB Device 3 Partition 3 Device 4 Partition 4 Device 5

Scalability Considerations This is a bread-and-butter serverless scenario – elastic growth Process messages in a batch (`EventData[]`) – “cardinality: many” Only one active instance per partition You can guarantee ordering with Event Hubs You must handle exceptions yourself. No deadletter queue Messaging choices: Event Hubs – high scale, high throughput, ordering is preserved Service Bus Queues / Topics – ”never lose a message”, ordering via sessions. Good for high variable workloads Storage Queues – inexpensive, less “enterprise” and message broker features Event Grid – send an ”event” out, any subscribers can get poked to be alerted. No ordering, no transnationality

TIP 3 - EventHub Scale Options Add partitions You can raise beyond 32 by contacting support 🤫 Increase Functions instance size Premium Plan

Inner and Outer Loop Development Build and Test (CI) Deploy (CD) Run (Prod) Inner Loop Code Run Debug Source Control Monitoring and Diagnostics

Inner and Outer Loop Development Build and Test (CI) Deploy (CD) Run (Prod) Inner Loop Code Run Debug Source Control Monitoring and Diagnostics

Tip 4: Azure DevOps for CI/CD

Dev Ops GA of Functions Build task New streamlined CLI command 11/17/2019 2:03 PM Dev Ops GA of Functions Build task Easily add Functions to a CI/CD pipeline New streamlined CLI command az functionapp devops-pipeline create Automatically configures DevOps to build with new commits to your version control Configures Github or Azure Repos automatically aka.ms/functions-azure-devops https://docs.microsoft.com/en-us/azure/azure-functions/functions-run-local#v2 https://docs.microsoft.com/en-us/azure/azure-functions/functions-create-first-azure-function-azure-cli https://stackoverflow.com/questions/46877667/how-to-push-a-new-initial-project-to-github-using-vs-code © Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

Inner and Outer Loop Development Build and Test (CI) Deploy (CD) Run (Prod) Inner Loop Code Run Debug Source Control Monitoring and Diagnostics

Tip 5: Monitoring Gain real-time observability Analyze and debug traces and metrics View dependencies and relationships with AppMap Application Insights Optional: demo of app map © Microsoft Corporation. All rights reserved.

Application Map

Distributed Tracing

Alex Karcher 5 tips for production ready Azure Functions Product manager – Microsoft 5 tips for production ready Azure Functions