Docker serverless v1.0
1 like7,545 views
Serverless computing is a cloud-native paradigm where developers build and run applications without managing infrastructure. It involves short-running, stateless functions that are triggered by events. With serverless, applications automatically scale up or down based on usage, and customers only pay for the compute time used. The document discusses serverless offerings from various cloud providers, demos serverless architectures using Docker containers, and notes serverless is well-suited for event-driven workloads like mobile backends and IoT but not long-running stateful processes.
![What is Serverless Computing ?
Is a cloud-native platform & app design paradigm
( which DOES involves Servers which are ‘invisible’ to developers [& Ops] ! )
for
short-running, stateless computing
and
(primarily) event-driven applications
which
aims to radically simplify development and operations
and
scales up and down instantly and automatically at runtime
and
billed for actual usage of compute time ( and/or memory)
at millisecond granularity
Serverless == React Compute == FaaS == No Servers To Manage (*)](https://image.slidesharecdn.com/dockerserverlessv1-170107121706/85/Docker-serverless-v1-0-2-320.jpg)
Ad
More Related Content
What's hot (20)
Similar to Docker serverless v1.0 (20)
Ad
Recently uploaded (20)
Ad
Docker serverless v1.0
- 1. Serverless Computing & Docker 07-Jan-2017, Docker Bangalore Meetup @Go-Jek Thomas Chacko thomasch (at) iconnexions.com
- 2. What is Serverless Computing ? Is a cloud-native platform & app design paradigm ( which DOES involves Servers which are ‘invisible’ to developers [& Ops] ! ) for short-running, stateless computing and (primarily) event-driven applications which aims to radically simplify development and operations and scales up and down instantly and automatically at runtime and billed for actual usage of compute time ( and/or memory) at millisecond granularity Serverless == React Compute == FaaS == No Servers To Manage (*)
- 3. source : martinfowler.com Monolith Application μ Srv μ Srv μ Srv f(x) f(x) f(x) f(x) Tier-App Tier-App Tier-App Serverless Architecture
- 4. What is Serverless architecture suitable for ? Serverless is suitable for short-running, stateless, event-driven workload Serverless is NOT suitable for long-running, stateful, number crunching workload Microservices Transactional databases Mobile Backends Deep learning/ML training Bots, ML interfacing Heavy-duty stream analytics IoT Spark/Hadoop analytics jobs Data (Stream,Log) processing Numerical Simulations Image processing, Cognitive Video Streaming
- 5. Why is Serverless gaining popularity? Efficiency Costs Complexity Developers : Simple coding - bits of code Enhanced scalability Focus on feature and UE - deliver value ( abstract away Infrastructure ) Accelerate MVP, TTM IT/Ops : Pay by millseconds - pay per execution model Optimisation of resource Gain infrastructure utilisation (time sliced)
- 6. Serverless offerrings Vendor / Framework Maturity Support for functions coded in Amazon Lambda GA Node.js , Python, Java , C# Google Cloud Function Alpha Node.js Microsoft Azure function GA - v1.0 JS, Python, C#, PHP, scripting bash, PS IBM OpenWhisk JS , Swift, Python, Java, Any (via Docker) iron.io ( Iron Function ) Alpha-2 Any language ( via Docker packaging ) Docker PoC Any language ( via Docker packaging )
- 7. IronFunction demo • Fire up an Iron/function containter $ docker run --rm -it --name functions --privileged -v ${pwd}/data:/app/data -p 8080:8080 iron/functions • Write a function e.g. func.go • create func.yaml file, replace $USERNAME with your Docker Hub username. $ fn init $USERNAME/hello • Build, test the function and if ok - push to docker hub $ fn build $ fn run $ fn push • Create an app ‘myapp’ and set a route to map /hello to function. An App is a grouping of function == API gateway $ fn apps create myapp $ fn routes create myapp /hello • Call the function from ‘anywhere’ . Each APP has its own namespace $ curl http://localhost:8080/r/myapp/hello ; curl $ fn call myapp/hello ; CLI $ echo '{"name":"All Ye Docker Enthusiasts"}' | fn call myapp /hello ; parameter pass at runtime • Supports async function calls too for CPU heavy tasks like image/data/video processing. Non-blocking as async function calls go in message queue and executed when resource become eventually available.
- 8. Demo 1 : Serverless App with Docker voting app msg queue worker database result app entry point database handle vote HTTP request handle result HTTP request process vote tasks continous runnng services continous runnng servicesf(x) container on demand source : Ben Firshman donkerconf 2016 demo backends : worker & message queue replaced with Docker container run on demand ( practically should scale to swarm ‘size’ on demand ) web frontends : voting app , result app replaced with Docker containers which spins up serving single HTTP request -> triggerred by a light weight HTTP server (entry point ~ LB ) ( practical ?? latency of container cold start ~ 300ms - 1 s ) http://node-ip/vote http://node-ip/result
- 9. Demo 2 : Using Funker on Swarm mode voting app database result app process_vote continous runnng services continous runnng servicesf(x) container on demand Funker (package): Function as Docker container Define function as services When initiated, opens a TCP port and wait for connection To call function - another docker service connects to function at its hostname (any node within swarm) process_vote/handler.js : var funker = require('funker'); funker.handle(function(args, callback) { // Return immediately so this fx runs in background callback(); .. process vote ... } vote/app.py funker.call("process-vote", voter_id=voter_id, vote=vote) node-ip:5000 node-ip:5001
- 10. One perspective of where we are headed with this…