What is serverless computing?
• Machine as the unit of scale
• Abstracts the hardware
• Application as the unit of scale
• Abstracts the OS
• Functions as the unit of scale
• Abstracts the language runtime
How do I choose?
• “I want to configure machines,
storage, networking, and my OS”
• “I want to run servers, configure
applications, and control scaling”
• “Run my code when it’s needed”
Microservices and AWS Lambda
AWS Lambda + Amazon API Gateway is the
easiest way to create microservices
• Event handlers one function per event type
• Serverless back ends one function per API / path
• Data processing one function per data type
AWS Lambda: Serverless computing
Run code without servers. Pay only for the compute time you consume. Be happy.
Triggered by events or called from APIs:
• PUT to an Amazon S3 bucket
• Updates to Amazon DynamoDB table
• Call to an Amazon API Gateway endpoint
• Mobile app back-end call
• And many more…
Makes it easy to:
• Perform real-time data processing
• Build scalable back-end services
• Glue and choreograph systems
No servers to
Never pay for idle
– no cold servers
Benefits of AWS Lambda
• Buy compute time in
100 ms increments
• Low request charge
• No hourly, daily, or
• No per-device fees
Never pay for idle!
1 million requests and 400,000 GBs of compute
every month, every customer
Using AWS Lambda
Bring your own code
• Node.js, Java, Python
• Bring your own libraries
(even native ones)
Simple resource model
• Select power rating from
128 MB to 1.5 GB
• CPU and network allocated
• Reports actual usage
• Securely grant access to
resources, including VPCs
• Fine-grained control over
who can call your functions
• Call or send events
• Integrated with other AWS
• Build whole serverless
Using AWS Lambda
• AWS SDK built in (Python
• Lambda is the front end
• Use processes, threads,
/tmp, sockets normally
• Persist data using Amazon
DynamoDB, S3, or
• No affinity to infrastructure
(can’t “log in to the box”)
• Author directly using the
console WYSIWYG editor
• Package code as a .zip and
upload to Lambda or S3
• Plugins for Eclipse and
• Command line tools
Monitoring and logging
• Built-in metrics for requests,
errors, latency, and throttles
• Built-in logs in Amazon
But what *is* AWS Lambda?
Linux containers as an implementation, not a programming
or deployment abstraction
• Process and network isolation, cgroups, seccomp, …
Predictive capacity management
• Purpose-built, massively scaled language runtime delivery
Swagger interpreter (API Gateway)
Amazon API Gateway: Serverless APIs
Any other publicly
Benefits of Amazon API Gateway
Create a unified API
front end for
throttling for back-
AWS Lambda, API Gateway regions
Lambda: Use cases
Data processing Back ends
Use case: Data processing
Example: Amazon S3 bucket triggers
Amazon S3 bucket events
Why functions are the right answer
Use case: Automatically scalable back ends
1. AWS Mobile SDK + Amazon Cognito for mobile app
Or AWS IoT for devices
2. Amazon API Gateway (if you want your own endpoint)
3. AWS Lambda runs the code
4. Amazon DynamoDB holds the data
Use case: Serverless web app architecture
1. Amazon S3 for serving static content
2. AWS Lambda for dynamic content
3. Amazon API Gateway for https access
4. Amazon DynamoDB for NoSQL data storage
in AWS Lambda
Data stored in
API GatewayStatic content in
Use case: New app ecosystems:
Alexa apps + Slack = serverless bots!
Alexa, tell Slack to
send, “I’m giving the
Message retrieval through scheduled
“Break a leg!”
(via Slack API)
22Commercial in confidence
23Commercial in confidence
eBuilder’s Effortless Device Care solution creates effortless
device self-service for mobile operator customers!
Device Care and an
end-user with a
What is appealing to us in Lambdas?
#1 Delegate more to AWS - Simplify deployment and management
#2 Cost efficiency - pay for actual usage not for statically allocated resources
#3 Functional approach - Helps promote well-structured system design
What are our use cases for Lambdas?
api gw lambdas
What are our experiences so far?
create full-stack developers
#2 Some activation latency to consider (we have seen 2s at worst) -
operationally very effective:
(latency of 394ms @ 95% for API-gw + 3 Lambdas + Dynamo)
#3 Lots of activity around automation - we started with Apex, now using
#5 Nice fit with Domain Driven Design (DDD): Lambdas + Kinesis streams
#1 Insufficient data as of now to compare operational costs with static EC2:s
• Scheduled functions
• Longer running times (5 min.)
• Higher code storage limits
(from 5 GB to 75 GB)
• New region: Frankfurt
• Node.js 4.3.2
• Swagger API import
• 1-minute schedules
• AWS CloudFormation support
for API Gateway.
Function schedules: The how-to guide
How can I keep a function warm (no cold starts)?
How can I poll a queue (like SQS)?
Schedule a function to read the queue.
How can I get more timers?
Have one scheduled function async invoke other functions.
How can I get granularity finer than 1 minute?
Run a background timer in your scheduled function.
Function versioning: The how-to guide
How can I get mutable configuration info?
Read it (e.g. from DynamoDB) during function initialization.
Wrap your config in a function and call it from your published code.
How do I “roll back” in AWS Lambda?
Using aliases, just switch what the alias points to.
(As a collection, add API Gateway stages or CloudFormation.)
How do I do blue/green deployments?
AWS Lambda handles fleet deployments, but if you want to shape
traffic, put a second “traffic cop” function in front.
How can I lock a client/device onto an old version?
Point them directly to that version’s ARN.
AWS Lambda VPC basics
• All Lambda functions run in a VPC, all the time
• You never need to “turn on” security – it’s always on
• You can also grant Lambda functions access to resources in your own VPC
• How: Add VPC subnet IDs and security group IDs to the function config
• Typical uses: RDB, ElastiCache, private EC2 endpoints
• Allows access to peered VPCs, VPN endpoints, and private S3 endpoints
• Functions configured for VPC access lose internet access…
• unless you have managed NAT or a NAT instance in the VPC
• …Even if you have “Auto-assign Public IP” enabled
• …Even if you have an internet gateway set up in your VPC
• …Even if your security group allows all outbound traffic
AWS Lambda VPC feature: Best practices
VPC is optional – don’t turn in on unless you need it.
The ENIs used by Lambda’s VPC feature count against
Ensure you have enough to match your peak concurrency levels
(we’ll consolidate where we can).
DO NOT delete or rename these ENIs!
Ensure your subnets have enough IPs for those ENIs.
Specify at least one subnet in each Availability Zone
Otherwise, Lambda will obey, but can’t be as fault-tolerant.
The serverless compute manifesto
Functions are the unit of deployment and scaling.
No machines, VMs, or containers visible in the programming model.
Permanent storage lives elsewhere.
Scales per request. Users cannot over- or under-provision capacity.
Never pay for idle (no cold servers/containers or their costs).
Implicitly fault-tolerant because functions can run anywhere.
BYOC – Bring your own code.
Metrics and logging are a universal right.