AWS Summit 2019 - Session 3: AWS Lambda
June 17, 2019•424 words
serverless - basically how to get to just the unique business logic for you, any common stuff (OS etc) is configured away so you don't spend time on what other people are already doing
Updating Lambda?
- Heroku style pushing - https://seed.run/docs/
- AWS CLI?https://docs.aws.amazon.com/cli/latest/reference/lambda/index.html probably AWS cli every language tons of event sources to kick off lambda (url is most important) AWS CodeCommit could be the git hub of Lambda the Lambda event variable holds submitted data
Lambda Layers reference within any function - modules or library basically, since Lambda's can't include each other
Amazon API Gateway
- create unified http endpoint for multiple microservices
- DDoS protection and throttling
- look up SWAGGER apis
- Authenticate and authorize requests to backend
- throttle, meter and monetize API
AWS Step Function
- workflows of Lambdas
- it's a better SF Flow
- essentially an app builder
should learn more about IAM #todo
Cognito is a serverless user pool for authorizing access
CloudWatch - logging/monitoring
AWS X-Ray -- looks at performance, not sure where
bossDB at Johns Hopkins
Incredible
2 PB of data for 1mm cube of brain
- index of synapses
- image slices build up to voxels
users upload 16 image tiles, they are processed into voxels, uploaded to S3 and catalogued in DynamoDB
they ran into concurrent Lambda throttling and would lose things
added another queue to track tiles instead of/back up to Lambda
SQS is interesting #todo
they now have a smarter activity to check for lambda throttling
They use a dead letter queue to see what jobs have failed:
Q1 --> L1
| |
Q2
Now though the Event Source can fire multiple lambdas with limits, to prevent throttling
Lambda for Server Management
JH uses lambdas to monitor servers, then update DNS etc if one goes down
AWS Secrets manager
Think through microservices, what can be done intermittently
From Disco
Event Source is a big deal, all sorts of actions based on it, including time sequence
Lambda can be scheduled/looped or triggered from certain inputs, or URLed
tune your lambda by CPU or memory
App Load Balancer with Lambda - use Advanced Request Routing, like on version or whatever
Lambda Layers is pretty great
By default Lambda is all public on Amazon's own VPC
AWS AppSync is GraphQL with sugar - offline ability apparently
Using GraphQL can let it use multiple Resolvers - ex two DynoDB and LKambda, return as one result
can include security permissions on specific fields in GraphQL