- FaaS vs PaaS
- dead letter queue
- mocking, testing
- open source FaaS (google, Serverless: https://github.com/serverless/serverless)
- 12 factor app https://12factor.net/
- https://www.ibm.com/cloud/learn/faas
- FaaS is subset of Serverless
- Serverless includes compute, storage, database, messaging, api gateway
- FaaS is only event driven compute
- best practices
- single responsibility functions
- isolate function => don't directly call others, so use events which are asynchronous
- comparison to PaaS, containers, VM
- faster provisioning time, in area of milliseconds
- no/nearly none administration overhead
- state and persistent connections have to be kept somewhere else
- no os/runtime maintenance (PaaS too) compared to IaaS
- no idle resource, functions stop after running
- only platform with resource limits
- https://aws.amazon.com/lambda/
- use case
- process data at scale
- https://cloud.google.com/functions/
- Integrated security and access management
- Faster development cycle by writing and running small code snippets
- avoid lock-in with open source FaaS
- superset of FaaS
- database
- user management with option of social login
- integrated analytics
- https://www.heise.de/developer/artikel/Servicemodelle-in-der-Cloud-5033939.html
- complete vendor lock in
- little customization possible
- Datenschutz
- Database DBaaS
- Message Queue MQaaS
- Firecracker VM
- MicroVM
- https://firecracker-microvm.github.io/
- https://github.com/openfaas/faas
- https://www.openfaas.com/blog/template-store/
- 4 parts:
- Dockerfile: base image with runtime if needed
- Entrypoint: How the function gets called. Either binary over STDIO oder HTTP Process
- Handler: User defined component
- Package List: Dependencies
- Dockerfile
- fprocess Variable setzen mit Befehl zum programm starten
- 4 parts:
- https://blog.alexellis.io/announcing-function-store/