Joint work with Erik Wittern and Waldemar Hummer Josef Spillner

AWS Lambda and #serverless. What’s all the fuzz about?

Philipp Leitner

philipp.leitner@chalmers.se

@xLeitix

Chalmers !2

https://icet-lab.eu

@IcetLab

We are hiring a postdoc!!

Chalmers !3

Function-as-a-Service?

Chalmers !4

Function-as-a-Service? Serverless?

Chalmers !5

Source: https://aws.amazon.com/lambda/

Chalmers !6

Source: https://aws.amazon.com/lambda/

Chalmers !7

Source: https://aws.amazon.com/lambda/

Chalmers !8

Source: https://aws.amazon.com/lambda/

Chalmers !9

Source: https://github.com/aws-samples/lambda-refarch-streamprocessing

An Example: Real-Time Tweet Processing

Chalmers !10

Internally

Container Runtime

Events(e.g., client

requests, system events,

triggers…)

User/tenant datastore

Execution datastore

e e e…

Execution QueueResult data

Function Controller

Function datastore

Function source code

Function container

<fn>

Event Controller

…

Load balancer

Authentication

Rule engine

Authorization Function container

<fn>Function container

<fn>

Source: loosely based on IBM’s OpenWhisk architecture

Chalmers !11

No state

Chalmers !12

No state

Hard cap on max. execution time

Chalmers !13

No state

Hard cap on max. execution time

Extreme tail latency

Chalmers !14

Chalmers !15

Methodology

Interviews (n = 12)

Analysis of Grey Literature (n = 50)

Web Survey (n = 182)

Chalmers !16

Main Findings

Compositional Application Model

“Microservices on steroids”

Chalmers !17

Source: https://github.com/aws-samples/lambda-refarch-streamprocessing

Recall: Real-Time Tweet Processing

Chalmers !18

Compositional Application Model

Chalmers !19

Main Findings

Compositional Application Model

Use Cases

Chalmers !20

Main Findings

User-Facing

27 (2

9%)

Backend

29 (3

1%)

Both38 (40%)

But: many challenges in user-facing apps

Chalmers !21

Use Cases - Backend

Chalmers !22

Main Findings

Compositional Application Model

Use Cases Challenges

Chalmers !23

Challenges

Chalmers !24

The Future (as we see it)

“Faster horses” Different Servicesvs

(end user view) (provider view)

Chalmers !25

The Future (as we see it)

Different Services

“Function-as-a-Service as the assembly language of the cloud”

Chalmers !26

The Future (as we see it)

Different Services

“Function-as-a-Service as the assembly language of the cloud”

Chalmers !27

Learn More

https://peerj.com/preprints/27005/