Connecting 10000 cars to the AWS Cloud Ashwin @ Zoomcar | Oct-12, 2019 

$whoami • Old school Sys / Network Admin • <3 Shell / UNIX • <3 Cloud and scaling tech companies • IoT is my new tinker toy • Own Mail and DNS servers 

No content

Note: This works when copying entire slides from other presentations as long as the source presentation is also 16:9 India’s first and largest app-based self- drive car rental company. 

Cars E-Scooters 

Note: This works when copying entire slides from other presentations as long as the source presentation is also 16:9 100% AWS 

Some more data • 40 Cities • 3mn users on the platform • > 4000 rides every day • 5lac km travelled daily • 10000 cars 

Some more data contd… • All cars are connected to platform • > 500mn IoT events ingested daily • > 1TB data lake for analysis and reporting 

Note: This works when copying entire slides from other presentations as long as the source presentation is also 16:9 This is just cars! We have scooters too ;-) 

How did we do this? 

Note: This works when copying entire slides from other presentations as long as the source presentation is also 16:9 WE BURNT OURSELVES, OUR FAMILIES, OUR FRIENDS, OUR LAPTOPS and then some! 

THAT CHEAP Datacenter! 

• ~ 70 % uptime • No NOC • Noisy neighbor problems • Hands tied on architectural, design, infra decisions • DevOps === Red eyed devils on the floor • Maintain / own Systems • Maintain / own Infra 

Bring on the experts! 

• Mandate to scale – Zap subscribe • 1 month to create migration plan • 2 days to create instances • 3 days to migrate data • 4 hours of downtime • 5 Star Performance 

Post the upgrade • Cloudfront – Website (1.2 secs to first paint) • Autoscaling API cluster (2sec to 150ms avg resp time) • EKS (~30% cost advantage) • Multi AZ RDS (No DBAs in office) • Read Replica Slaves (Daisy chained on demand for teams) • AWS Glue / Athena / Redshift (More data sources, more information) • 40X average scale / 200X peak load (Yaaaaay!!!) • Elasticache (Memcache / Redis at their performant best) • Kafka for IoT pipelines (Beast mode on steroids for distributed data processing) 

Data is the new FUEL! 

Business Intelligence • User behavior capturing • Quicker identification of UX drop off points • Demand Forecasting • Predictive Maintenance 

Data Analytics • Every data point we can think of captured • Cars produce 500mn data points per day • Customer Churn • Fraud detection • Driver Score 

Note: This works when copying entire slides from other presentations as long as the source presentation is also 16:9 And that’s how we connect 10000 cars to the AWS Cloud… 

Note: This works when copying entire slides from other presentations as long as the source presentation is also 16:9 Hold on! There’s more! 

Note: This works when copying entire slides from other presentations as long as the source presentation is also 16:9 How would YOU do it? 

Note: This works when copying entire slides from other presentations as long as the source presentation is also 16:9 How would you connect 1mn devices? 

Note: This works when copying entire slides from other presentations as long as the source presentation is also 16:9 Few Key Learnings 

Bad arch decisions bite you hard 

Be prepared to change at the earliest 

Managed Services exist for a reason 

Note: This works when copying entire slides from other presentations as long as the source presentation is also 16:9 AWS IoT 

Note: This works when copying entire slides from other presentations as long as the source presentation is also 16:9 Key Components 

Ingestion / Persistence / Analysis 

Key Components • Applications • Device Management • Message management • Storage • Data Processing • Notifications • Analytics 

Key Components • Applications – AWS IoT SDK / Edge with Greengrass • Device Management – AWS IoT Hub / Device Shadows • Message management – AWS IoT Message Queue / AWS IoT Rules Engine • Storage – RDS / DynamoDB • Data Processing - Lambda • Notifications – SQS • Analytics – EMR / Kinesis / Redshift 

Why AWS IoT??? 

• Rock solid / proven ecosystem • Infinite Scale • Insanely small Time to Market • Programmability • Insights • Smooth learning curve • Existing support systems and community 

No content

Note: This works when copying entire slides from other presentations as long as the source presentation is also 16:9 Thank You [email protected] +91 9003010231