Serverless Computing and the Apache OpenWhisk Project

Transcription

1 Serverless Computing and the Apache OpenWhisk Project #openwhisk Matt Rutkowski, STSM, IBM Cloud Open Technologies 1

2 Session Topics What is Serverless computing? What is Apache OpenWhisk? OpenWhisk s Event-Driven Programming Model Apache OpenWhisk Project Overview and Architecture Apache OpenWhisk Community 2

3 What is Serverless Computing? No Servers, Just Code 3

4 The Evolution of Cloud Platforms HW SDS Cloud Platforms Platform-as-a-Service (PaaS)... Infrastructure-as-a-Service (IaaS)... Identity / Access Database OSS / BSS Compute Storage Network Evolution Cloud Datacenters Software-Defined-Services (SDS) abstraction of Hardware Compute, Network and Storage Hardware Fabric Services Security, Data, Operational, Business, etc. IaaS = Virtual Machines (VMs) Workloads: Heterogeneous applications Tend to be tightly-coupled, Monolithic Configure / Manage Stacks of applications Web/App Server, DB, logging, Complex Network & Storage Configurations IPs, Ports, Routes, SANs, Volumes, Plan / Predict for Scaling Needs CPUs, Memory, Load Balancing, Lots of Configuration & Pre-planning for worst case Scaling & Failover

5 The Evolution of Cloud Platforms Cloud Platforms HW SDS Platform-as-a-Service (PaaS)... Infrastructure-as-a-Service (IaaS)... Identity / Access Database OSS / BSS Compute Storage Network Evolution PaaS = Containers Workloads: Homogenous applications Utilize abstract Configuration, Scheduling and Management frameworks Kubernetes, Apache Mesos, Setup Monitoring tools for Container Apps Prometheus, Logstash, Configure Platform services interactions: Storage, Access Control, Key Mgmt., MicroServices (often run in Containers) Split large codebases into small, standalone services Single light weight, RESTful endpoints or APIs Protocol aware (HTTP, HTTPS) Built in groups or suites around resources Run inside MicroService Frameworks to handle Configuration, logging, metrics, service registration & discovery, protocols, etc. e.g., Flask (Python), Spring (Java), Gizmo (Go) Still Lots of Configuration to work with Frameworks and Resources

6 The Evolution of Cloud Platforms Serverless Computing is the next step for Cloud platforms Cloud Platforms Function-as-a-Service (FaaS) Platform-as-a-Service (PaaS) Infrastructure-as-a-Service (IaaS) HW SDS... Identity / Access Database OSS / BSS Compute Storage Network Evolution Serverless = Functions Workloads: Simple, single-tasked Functions No Back-end Servers Configuration Automatic scaling, based upon load Driven by events, and their data Majority of Functions & Orchestration are Front-end around workflows and tasks around the applications data where Developers Focus on ONLY writing Application and Business logic! There are still Servers! But they are a No-Op for you! Provider s DevOps teams configure, manage and assure Functions scale and run efficiently. No Configuration of Servers, Only Pay for Compute time functions actually use

7 Characteristics of Serverless Workloads: Data Functions are: Data Driven Event Input Results Output data often JSON All about Front-end resources What s a Server? What s a Network? Protocol (route) agnostic HTTP? What is that? Event-Driven Run only when associated Event Triggers them Serverless Apps try to: Maximize stateless or pure functions Process specific data & return results quickly Do not block / poll / wait for a a resource s state change (results) Optimistic that Input data is correct, valid. Minimize stateful interactions or impure functions Push state (results) to stateful services, for example Data Stores / Message Queues 7

8 What is Apache OpenWhisk? whisk (v) : to move nimbly and quickly. 8

9 Apache OpenWhisk is an open source Serverless platform Designed to connect a fast-moving, event-driven world IoT DevOps Mobile Analytic Cognitive Micro Services OpenWhisk provides a robust, scalable Serverless platform for rapidly integrating services and data from any domain using events that drive simple integration code. The name Whisk is also used for a kitchen utensil Developers, the Chefs, can use OpenWhisk to quickly compose compelling applications, 9 by mixing together interesting functions using event data

10 Many Cloud providers offer some form of Serverless framework AWS Lambda Azure Functions Cloud Functions Functions others Amazon MicroSoft Google Iron.io Brocade Apache OpenWhisk offers: Apache Software Foundation (ASF) True, community-driven open source (Apache 2 License) Proven on IBM s Bluemix Cloud Platform Exact, same code in open source Highly Performant with strong, automated integration testing < 10 millisecond schedule/load into runtime, results caching 10

11 History of OpenWhisk Early 2015 Feb 2016 IBM Research begins coding IBM debuts and open sources on GitHub (openwhisk.org) & announces Aug 2016 Nov 10, 2016 Adobe partners with IBM on API Gateway for OpenWhisk IBM Bluemix beta offering Nov 23, 2016 Dec 14, 2016 OpenWhisk Incubated at Apache Software Foundation General Availability (GA) on Bluemix Feb/Mar 2017 May 2017 MessageHub (Kafka), Web Actions, VS Code OpenWhisk s API Gateway integration GA

12 Apache OpenWhisk has many deployment options Local Development OpenWhisk Serverless Platform Windows All component services are built into Docker containers Distributed docker compose 12

13 Event-Driven Programming Model performing the correct amount of work at the time needed 13

14 Why an Event-Driven Programming Model? Everything produces Events in the Natural World and in IT 14

15 OpenWhisk s Programming Model A Action : A stateless, relatively short-running function invoked as an event handler. Input JSON Object Stateless Function A Output JSON Object Goal: An Action s run time is measured in milliseconds, well under the defaulted maximum of 5 minutes 15

16 OpenWhisk s Programming Model T Trigger : The named channel for a class of events. Triggers represent the events (and their data) themselves without any concept of how they were generated. Event sources T (logical, named channel) OpenWhisk Serverless Platform Note: In a pub-sub system, a trigger could be viewed as a message topic

17 OpenWhisk s Programming Model R Rule : A mapping from a Trigger to an Action which may contain simple conditional logic. T Trigger R (conditional logic) A Action Note: OpenWhisk evaluates incoming events (that belong to a Trigger) and invokes the assigned Action (event handler) associated by the Rule. 17

18 Programming Model F Feed : A Feed manages the stream of Events from an external Event Source with some optional control operations called Feed Actions : CREATE, DELETE, PAUSE, UNPAUSE The running Feed fires events to an associated Trigger Event source Feed Feed Actions T A CREATE DELETE Trigger Action PAUSE Note: Feeds can Fire Triggers without an Event Source such as the OpenWhisk Alarm package. 18

19 Programming Model P Package: A named, shared collection of (Namespaced) Actions, Triggers, Feeds, Rules, P A Package T R F manifest.yml Note: Designed as a Hirst-class entity in the OpenWhisk Model, being used by new Whisk Deploy tool ( wskdeploy ) for 1-click Deployments 19

20 Apache OpenWhisk Event-Trigger-Rule-Action Processing OpenWhisk Platform Sample Event sources Events Packages Package Triggers / Rules Actions & Action Sequences External Service Eco-System Data Stores Feed Rule Action NodeJS Social Media Package Feed Rule Action 1 NodeJS Action 2 Python Action 3 Swift 3rd Party or Self-Enabled REST APIs Message Queues Events Feeds Package Instrumented Machines Feed P F Rule T R Action 1 Swift A Action 2 Docker 20

21 Apache OpenWhisk Project Architecture & Overview 21

22 OpenWhisk Platform Architecture Client endpoints SSL Termination Router OpenWhisk Core OpenWhisk Controller Load Balancer Activator Message Bus 1 topic per-invoker Invoker Invoker Read message, Spawn Docker container language runtime (JS, Python, Swift, etc.) HTTP request / response Messages for Invoker OW CLI REST Key-Value Store Document Store OpenWhisk APIs (CRUD) manages the Kafka cluster, tracking status of nodes present in Kafka cluster topics, messages, and quotas. Docker Containers Source of Truth Deployment, Configuration Information Actions Triggers Rules Namespaces Credentials Metadata etc. Ansible Deployments OpenWhisk Platform Components are all Docker containers 22

23 Invokers Run Language-Specific Docker Containers for Actions OpenWhisk supports many languages (runtimes) for Actions JavaScript, Swift, Java, Python and more in the future A OpenWhisk Controller Automatically Schedules the correct language runtime container for your code Caches Action code keeping it warm Automatically monitors performance and scales containers OpenWhisk Container Scaling nodejs python swift java docker Clusters other IaaS Docker Engine Compute Nodes (Virtual Machines)

24 Apache OpenWhisk Eco-System Overview Tooling wskdeploy (deployer + spec.) debugger devtools (Compose, Kube) playground (XCode Extension) xcode (x3) vscode Command Line Interface CLI ( wsk ) Router OpenWhisk Core Key-Value Store API Gateway Controller Document Store Ansible Deployments Message Bus OpenWhisk Catalog Invokers Packages Alarms Kafka Cloudant Push Notifications JIRA RSS Application Samples slackbot slackinvite GitHubSlackBot matos External Resources Workshop, Tutorial Template openwhisk.github.io (openwhisk.org) 24

25 Core OpenWhisk Platform Repository OpenWhisk incubator-openwhisk Contains the OpenWhisk Serverless Platform code Scala is the Primary Language of the Controller Manages Whisk System Entities / State; Load balances Invoker requests; Provides Logging Ansible run books for various deployment environments i.e., Local (single Vagrant VM), Distributed, Mac System & Utility Catalog Repository Catalog incubator-openwhisk-catalog Catalog of built-in system, utility and sample Actions and Feeds Utilities: Combinators (forwarder, retry, etc.), Git, Slack, Weather, Web Socket, IBM Watson Samples: cat, countdown, curl, greeting, hello (variants), httpget, sort, split, trigger, wordcount, 25

26 Command Line Interface (CLI) Repositories Command Line Interface CLI ( wsk ) incubator-openwhisk-cli Implements wsk CLI using the Cobra framework Integrated Syntax Checking (POSIX flags) / Auto-completion / Validation / Help Supports Go API Client as a plug-in Supports client SDKs (component = docker swift 3 ios 9 &10 OSX) incubator-openwhisk-client-go Implements HTTP-REST APIs for OpenWhisk Actions. Triggers, Rules, Activations, Packages, Namespaces, Api Gateway, SDKs, more Http API Clients for other languages (Python, Swift, JavaScript) incubator-openwhisk-client-python incubator-openwhisk-client-swift incubator-openwhisk-client-js 26

27 Package Repositories Packages openwhisk-package-alarms Alarm (Feed) that fires a (trigger) event at a specified frequency (UNIX crontab ). openwhisk-package-kafka Package for communicating (Producer-Consumer) with Kafka (or IBM Message Hub) instances openwhisk-package-push-notifications Push Service: Google Cloud Messaging (GCM), Apple Push Notification (APNs), Firefox/Chrome/ Safari openwhisk-package-jira Integration with JIRA events. Simple issue creation supported openwhisk-package-rss Subscribe to RSS/ATOM feeds and receive events when a new feed item is available. openwhisk-package-cloudant Enables interaction (read / write / monitor) with a Cloudant database 27

28 API Gateway Repository API Gateway incubator-openwhisk-apigateway Main configuration files and a Dockerfile to build all modules into a single container. leverages technologies like Nginx, OpenAPI (Swagger), LUA Modules (plug-ins) Combined with OpenWhisk it provides an easy way to build back-ends without the need to manage servers by mapping APIs to actions Now supported via Whisk CLI using the api command (no longer api-experimental ) Recent Activities Management interface to configure actions as APIs OAuth2 support CORS support Profiling tools used to measure performance 28

29 Tooling Repositories Tooling wskdeploy (deployer + spec.) debugger devtools (Compose, Kube) playground (XCode Extension) xcode (x3) vscode incubator-openwhisk-wskdeploy OpenWhisk deployment tool wskdeploy, 1-click deployments Deploys OpenWhiske Packages ( manifest.yml files ) incubator-openwhisk-devtools Experimental deployments for testing OpenWhisk functions locally. using Docker Compose, NodeJS incubator-openwhisk-deploy-kube Kube deployment on Ubuntu (localhost) - as of May 16th All Docker images now shared between both Compose and Kubernetes incubator-openwhisk-debugger Debug Actions locally using the wskdb command using NodeJS (NPM package) incubator-openwhisk -xcode / -playground collection of OpenWhisk tools for OS X implemented in Swift 3 incubator-openwhisk-vscode Prototype extension for Visual Studio Code 29

30 Apache OpenWhisk Community 30

31 OpenWhisk Project Vital Statistics GitHub ( Core Repo.) Stars 1493: and growing ~10-20 per week on average Stats 311 forks Contribution Graphs: Contributors (all repos.) plans to formalize releases (and binary downloads) from Apache so we can start tracking users 31

32 OpenWhisk Become a Contributor Find all OpenWhisk Repositories: More Integrations! (Queues, Devices, Data Stores) More Language Runtimes My personal More Packages! (Twilio, Facebook, Twitch, etc.) More Samples (Workflows of 3+ Actions) Verify / Fix / Create / Improve Documentation Whisk Deploy (wskdeploy) Package Manifests for all Curated Packages / Samples Web User Interface (Compose, Deploy, Debug via UI) More Web Actions (Mime Types) Whisk Package Catalog (NPM-like Distributed Registry) 32

33 OpenWhisk.org Connect with us! Latest Events & Meetups! Links to YouTube: Channel OpenWhisk SlideShare: Twitter: #openwhisk Slack: Invite yourself using a Whisk Action! Medium (blogs): StackOverflow: tag openwhisk 33

34 Thank you! Questions? 34