Cloud Computing: Introduction to Serverless

• The presentation is titled "Cloud Computing: Introduction to Serverless Computing".

What is Serverless Computing?

• With IaaS (Infrastructure as a Service), virtual servers are provided in the cloud, but they are bare machines
• PaaS (Platform as a Service) provides virtual application platforms, eliminating the need to manage the OS and manage scalability
• PaaS does not automate running numerous microservices and requires container orchestration platforms
• Serverless Computing provides compute resource abstraction to its logical conclusion

Evolution of Computing

• Computing has evolved to increase the focus on business logic and decrease the concern over stack implementation
• The progression is from Bare Metal -> Virtual Machines -> Containers -> Functions

What is Serverless?

• There is no one clear definition of Serverless
• In the first case (BaaS):
  • Serverless describes applications that use third-party, cloud-hosted applications and services to manage server-side logic and state
  • It uses "rich client" applications which use cloud-accessible databases and authentication services
  • These services are "(Mobile) Backend as a Service" (BaaS)
• In the second case (FaaS):
  • Serverless means applications where server-side logic is written but run in stateless compute containers that are event-triggered
  • These containers are ephemeral and fully managed by a third party
  • One way to think of this is "Functions as a Service" (FaaS)

NIST Definition of PaaS

• PaaS means "the capability provided to the consumer is to deploy onto the cloud infrastructure consumer-created or acquired applications created using programming languages and tools supported by the provider"
• The consumer manages the applications and not the cloud infrastructure

Serverless FaaS

• Serverless FaaS removes user control over hosting to provide simpler scaling and a more attractive billing model

Serverless Architecture

• In a serverless architectural system, the Edge consists of a UI, API Gateway and Cloud Event Sources
• These connect to the Master, which consists of an Event Queue and Dispatcher
• The Dispatcher connects to Workers

UI-Driven Applications

• Traditional three-tier client-oriented systems use server-side logic
• In this system, Logic in the system (authentication, page navigation, searching, transactions) implemented by the server application
• Serverless Architecture features a service for Authentication, direct access to a Database and split purchase/search functionality
• FaaS functions respond to HTTP requests instead of having an always-running server

No Centralized Management

• In a Serverless architecture there is no central arbiter, instead there is more choreography over orchestration
• It shares principles with a microservices approach:
  • Flexibility
  • Extendibility
  • Division of concerns
  • Cost reduction
  • Higher complexity

Example: Message-Driven Application

• Message-Driven applications are for backend data-processing services
• Serverless versions replace the long-lived message-consumer application with a "click processor" FaaS function
• The architecture becomes smaller and the system becomes event driven

What is FaaS (According to Amazon)

• "AWS Lambda lets you run code without provisioning or managing servers"
• "With Lambda, you can run code for virtually any type of application or backend service - all with zero administration"
• "Just upload your code and Lambda takes care of everything required to run"
• "and scale your code with high availability"
• "You can set up your code to automatically trigger from other AWS services"
• "or call it directly from any web or mobile app"

AWS Lambda

• AWS Lambda allows the user to "run code without provisioning or managing servers."
• FaaS is about running backend code without managing your own server systems or your own long-lived server applications
• A provisioned server and an always-running application are unnecessary

Subsecond Metering

• With AWS Lambda, "You pay only for the compute time you consume - there is no charge when your code is not running."
• So, one is charged for every 100ms the code executes and the number of times the code is triggered

Language Agnostic

• According to AWS, "With Lambda, you can run code for virtually any type of application or backend service - all with zero administration."
• FaaS offerings do not require coding to a specific framework or library
• FaaS functions are regular applications when it comes to language and environment
• AWS Lambda functions can be implemented in Javascript, Python, Go, any JVM language (Java, Clojure, Scala, etc.), or any .NET language or any language that can compile down to a Unix process

Resource Provisioning

• According to AWS, "Just upload your code and Lambda takes care of everything required to run."
• Deployment is very different from traditional systems
• In a FaaS environment we upload the code for our function to the FaaS provider, and the provider does everything else

Automatic Scalability

• According to AWS, "and scale your code with high availability."
• Horizontal scaling is completely automatic, elastic, and managed by the provider
• The vendor handles all underlying resource provisioning and allocation–no cluster or VM management is required by the user at all

Events

• Functions in FaaS are typically triggered by event types defined by the provider
• Amazon AWS includes S3 (file/object) updates, time (scheduled tasks), and messages added to a message bus

HTTP Requests

• "Or call it directly from any web or mobile app"
• Most providers allow functions to trigger as a response to HTTP requests by using an API gateway

Stateless

• FaaS functions have restrictions when it comes to local state stored in variables in memory, or in a local disk
• There is no guarantee that such state is persisted across multiple invocations
• FaaS functions are described as stateless, so any required consistent state must be externalized outside of its instance

Function Duration

• FaaS functions are usually limited in how long they are allowed to run
• AWS Lambda functions run for five minutes at most
• Long-lived tasks are not suited to FaaS functions

Startup Latency

• It takes some time for a FaaS platform to initialize an instance of a function before each event, from a few milliseconds to several seconds
• Initialization of a Lambda function include either:
  • Warm Start, reusing an instance of a Lambda function and its host container from a previous event
  • Cold Start, creating a new container instance, starting the function host process, etc
• Cold start may depend on the programming language, how many libraries are used, how much code there is, and the configuration of the Lambda function environment

Function Life-Cycle

• Check if an instance is running
• Either activate the process, or start an instance
• Handle the request and stop the process

API Gateway

• An API gateway is an HTTP server where routes and endpoints are defined in configuration
• Each route is associated with a resource to handle that route
• Handlers are often FaaS functions in Serverless architecture
• When an API gateway receives a request, it finds the routing configuration matching the request.
• In the case of a FaaS backed route, it will call the relevant FaaS function
• The FaaS function executes its logic and returns a result to the API gateway
• This result is transformed into an HTTP response that is given to the original caller
• API gateways may also perform authentication, input validation, response code mapping, and more
• Amazon's API Gateway is a BaaS service that does not need to be provisioned

Serverless Computing Vendors

• Amazon has the most recognized Serverless Computing offering, with AWS Lambda
• Other vendors include Google Cloud Functions, Azure Functions, IBM Cloud Functions, and Pivotal Function Service
• All of them enable you to deploy functions that are executed on-demand and billed by the millisecond of execution time
• Open source Serverless projects include Kubeless, OpenFaaS, and Fn Project

Open Source in the FaaS World

• Serverless Framework: aims to make easier use of tooling provided by AWS and also cross-vendor abstraction.
• AWS's own deployment tool, SAM(Serverless Application Model)is also open source.
• Apache OpenWhisk is an open source FaaS platform
• Microsoft open sourced much of its Azure Functions platform

Comparison with PaaS

• PaaS applications bring entire applications up and down in response to an event
• FaaS platforms do exactly this.
• FaaS and PaaS are different in scaling
• FaaS scales automatically, transparently, and finely.
• This is tied in with the automatic resource provisioning and allocation

Benefits of Serverless Computing

• Reduced operational cost (Serverless is, at its most simple, an outsourcing solution)
• Reduced development cost (A BaaS database removes much of the database administration overhead)
• Reduced packaging and deployment complexity
• Reduced time to market and high flexibility
• Reduced execution costs (save money using FaaS unless your server use is near its limits)
• Greener computing (Reduce inefficiency)

Benefits at a Glance

• Serverless provides milliseconds of provisioning time and blocks of milliseconds for charging granularity
• It also has blocks of milliseconds of costs
• Virtual Machines provide minutes of provisioning time and hours for charging granularity
• Containers provide seconds/minutes of provisioning time and minutes for charging granularity
• On Premise provides minutes of provisioning time and fixed costs

Drawbacks of Serverless Computing

• Vendor control and forced API upgrades may be an issue
• There may be multitenancy problems, vendor lock-in, execution duration, startup latency, and difficult testing
• Security, robustness and performance problems may also occur

Open Issues

• Minimizing the cold start problem
• Degree to automatically decompose legacy code
• Applications may fail without supporting stateful serverless applications with QoS
• Service relies such as identity providers may be unavailable
• Support for events that can be generated at the edge through increased adoption of IoT and mobile devices

When is Serverless Good?

• When it is short-running, stateless or event-driven
• Use for Microservices, Mobile Backends, Bots, ML Inferencing, IoT, Modest Stream Processing and Service Integration

When is Serverless not Good?

• When it is long-running, stateful or requires number crunching
• Use for Databases, Deep Learning Training, Heavy-Duty Stream Analytics, Spark/Hadoop Analytics, Numerical Simulation and Video Streaming

This presentation introduces serverless computing and its evolution from bare metal to virtual machines to containers to functions. Serverless computing abstracts compute resources, increasing focus on business logic and decreasing concern over stack implementation. It uses third-party, cloud-hosted applications.