A Heavy AWS User’s Firstlook at Azure I’ve used too many tools: • AWS, Azure • GCP, Alibaba, Tencent • Algolia, Stripe, Auth0, Vercel, … Work: • I help IoT folks leverage the

Lead Cloud Architect .com/in/fmc-sea independence Specific IP Integration OS, communication, connectivity BSP, Driver, Firmware Driver, BSP, Hardware Understanding Device connectivity Cloud Cloud Cloud integration devices Other Connectivity M2M smartphone & Desktop & mobile application Web, desktop

| 3 W I T E K I O | W e h e l p h i g h t e h m a k e r s b u i l d g r e a t s o f t w a re | 4 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re Summary

1. Introduction

2. Compute

3. Storage

4. IoT

5. Cloud Cost Optimization

6. Questions 1. Introduction What’s the Big Deal about Picking a Cloud?a Dealabout Picking Big the What’s • • Huge commitment for new companies new Huge commitmentfor critical for software businesses software for critical be set can feature and cost Infrastructure

| 7 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re Mythbusting use the cloud. : We need to to need We

| 8 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re Do you? Do We’re not Dropboxnot We’re though…

| 9 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re Building without “A Big Cloud”

| 10 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re Mythbusting use the cloud. : We can’t We

| 11 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re Azure Compliances: Azure

| 12 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re AWS Compliances: AWS

| 13 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re Mythbusting Moving to the cloud will : save money. save

| 14 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re Cloud resources can be spun can spun be resources Cloud People need eat. to up and down. up

| 15 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re Mythbusting We need to be multi : - cloud!

| 16 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re What does “Multi does What “We can move our business between between business our move can “We “We run some apps in Azure and inAzure apps some run “We AWS and Azure atany time” Azure and AWS some inAWS” some vs. - cloud” mean?cloud”

| 17 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re “We can move our business between between business our move can “We AWS and Azure atany time” Azure and AWS

| 18 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re with either either with folks hire can we twomeans clouds Using When CloudA CloudA has an outage we’ll be ready be outage hasan we’ll or or CloudB experience! Ooops two clouds exceptionally well twoexceptionally clouds Our team now needs to learn tolearn now needs Our team . DNS was only in in wasonly DNS . CloudA

| 19 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re Mythbusting We need to worry about : lock - in.

| 20 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re notice moments a at cloud toany switch can we so containers use toVMsonly and need We more way paying up end will we onecloud in too invest Ifheavily we two clouds will be more costly more be twowill clouds services will make developing developing make will services Writing your applications for applications your Writing Never leveraging managed managed leveraging Never applications much harder much applications

| 21 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re o o o Top Two Public Clouds Public Two Top andAzure? AWS Why 2020 2019 Gartner 2018 ZDnet ZDnet - - #1: AWS, #2: Azure #1: AWS, #2: Azure - #1: AWS, #2: Azure o o Other Options Other Cloud ProvidersOther Cloud (frequently GoogleCloud inthird)

| 22 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re Why AWS andAzure? AWS Why Azure IoT Azure

| 23 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re 2. Compute https://specify.io/concepts/serverless - baas - faas

| 25 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re • • • IaaS Networking services Networking EC2 Amazon Virtual Machines Azure The Compute Infrastructure Spectrum • • PaaS Azure App Services Azure Beanstalk AWS Elastic • Serverless Compute: • • Azure Functions Azure AWS Lambda

| 26 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re AWS AWS Azure Azure Cloud GeneralPurpose VirtualMachines A2 v2 Instance Type t3a. a1.large B2S IaaS Head to Head: Head: to Head IaaS *This is a tiny subset of all the instance types either cloud offers cloud either instancetypes the all tinysubset a of is *This 2 2 2 2 vCPUs 4 GiB 4 GiB 4 GiB 4 GiB RAM $0.0376 $0.051 $0.0416 $0.076 Hourly Price

| 27 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re What doesn’t this address? doesn’t this What instance typesinstance differences in Detailed

| 28 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re What doesn’t this address? doesn’t this What Windows licensing with Azure

| 29 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re from and why is my bill $200?” bill my is why and from resourcesthese come all did “Where Elastic Beanstalk Elastic PaaS Comparison: Azure App Service

| 30 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re Bonus! CaaS (Containers ) Service Elastic Service Container Elastic Azure Container Container Azure Azure Kubernetes Azure Service Service

| 31 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re AWS Lambda FaaS Growth Azure Functions Azure

| 32 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re AWS SAM AWS Lambda FaaS Development Frameworks Development Azure Functions Azure

| 33 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re HTTP HTTP FaaS Capabilities:

| 34 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re FaaS Capabilities Capabilities – Data Processing Data

| 35 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re • • • gumption(s?) some need still Functions Azure Pass on AWS PaaS Costs Compute Comparable Compute Takeaways Compute

| 36 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re 3. Storage • • • IaaS databases Host own your Amazon EC2 Machines Virtual Azure The Storage Infrastructure Spectrum • • • PaaS Azure DatabasesAzure Amazon Redshift Database Service AWS Relational • • • • Managed Services Azure Blob StorageBlob Azure DB Cosmos Azure Amazon DynamoDB

| 38 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re IaaS Database Comparison: Database IaaS Do you really want to host your your host to want really Do you own database servers? database own

| 39 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re Amazon Aurora Amazon Amazon RDS Amazon PaaS SQL Databases Azure Databases Azure Azure SQL

| 40 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re Purely HTTP Interfaces and SDKs and Interfaces HTTP Purely Design constraints to provide toconstraintsprovide Design consistentlatency DynamoDB Managed ‘NoSQL’ Databases Flexibility to leverage multiple multiple toleverage Flexibility APIs and business and APIsuse cases (including SQL (including Multiple interfaces Multiple CosmosDB – traitor!)

| 41 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re https://www.msp360.com/resources/blog/microsoft Azure Storage Accounts Storage Azure - azure - storage - types - explained/

| 42 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re Primarily HTTP Interfaces HTTP Primarily Simple Storage Service Storage Simple Simple Queue Simple Service DynamoDB, DynamoDB, AWS Storage Services Storage AWS Elastic Block Store (EBS)Store Block Elastic Elastic (EFS) System File Elastic Storage for Virtualfor Storage Machines

| 43 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re • • • multipleAWSservices with correlate Accounts Storage Azure Managed storage is the standard Learnloveto manageddatabases Storage Takeaways

| 44 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re 4. IoT • • • • Other mechanisms Attestation TPM Provisioning Unique X.509 Certificates SymmetricKeys (Azure only) • • • • JSON Web Tokens() Web JSON Preregistering device ids and public keys keys device with CSRs Device claiming (Azure Sphere) the Cloud Authenticating Devices to 46

| W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re AWS Architecture: Device Fleet Analytics Fleet Device Architecture: AWS

| 47 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re Comparable Azure Architecture Azure Comparable https://docs.microsoft.com/en - us/azure/architecture/reference - architectures/_images/iot.png

| 48 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re Bonus: Comparable GCP Architecture https://cloud.google.com/iot - core

| 49 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re 5. Cloud Cost Optimization • • sizing Right capacity unitscapacity Appropriate sizes instance Appropriate GeneralCloud Cost Optimization • • • Options Purchasing capacity Reserved instances Reserved Spot instances • • • Utilization balancing Load Auto scaling services Managed • • Buffering queues Buffer data in processing Stream data

| 51 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re • • • (AWS) Pricing Core IoT • • • (Azure) Tiers Hub IoT utilization Optimize applications to reduce features specific of utilization per Pay request per Pay 60 $25 Units S1 Tier Standard $10 Units Tier Basic - 80% cost reduction Specific Optimization Examples - $500/ mo - $2500/ mo • • • Blob Storage Access (Azure) Tiers Storage Blob • • ClassesS3 Storage (AWS) policies Creating appropriate lifecycle Archive Deep Glacier storage Standard policies Creating appropriate lifecycle Archive Cool, Hot, Premium, - $0.023/GB - $0.00099/GB

| 52 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re • • • • Azure IoT Hub IoT Azure Optimization Cost IoT AWS and Azure Tune device applications to reduce usage reduce to applications device Tune auto Setup units Hub IoT of type and number the Tune standard) vs. ( need you tier the Select • Do you need bidirectional communication or not? communication bidirectional Do you need - scaling for IoT Hub for scaling Azure IoT Azure • • • • AWS IoT Core IoT AWS Optimize message sizes for metering Reduce connectivity, messages, bundle data Reduce usage to only required features pricing feature and request per Pay • 8 KB message, 5 KB metering, charged at 10 KB

| 53 W I T E K I O | W e h e l p h i g h t e c h m a k e r s b u i l d g r e a t s o f t w a re 6. Questions? CONTACT ME CONTACT SALES [email protected] /in/fmc-sea witekio.com [email protected]