Serverless Days - London - My Key Takeaways

Twitter: https://twitter.com/serverlessldn

Official Announcement: https://london.serverlessdays.io/news/keynote-speakers-2019/

Table of contents:

• The what, why and who of the event
• Followed Agenda
• My Key TakeAways from the event

The what, why and who of the event

What: Serverless Days in London I found to be a great opportunity to see what other people are doing with the serverless technologies out there and what is the state of this space, in particular what is the state of the conference space and what level of maturity the community has in London.

WHY: I have been using serverless for more and more projects recently and wanted to see if some of the questions I have been running into are being answered by anyone else, I will be diving into some in the next chapter, luckily for me this conference has provided above and being my stated reasons to attend.

WHO: Besides the speakers, which I only have heard before briefly, the sponsors panel looked really promising as the sponsors are keeping a close eye on the space and seem to be quite close to the event either by organising workshops, doing their own talks or organising a boot.

Followed Agenda

The official event agenda can be found at serverless days - agenda

Observability workshop

I was fortunate enough to be able to get to one of the workshops. The workshop was titled Building a 100% Serverless Blog Site Application with the Right Observability, while aimed at beginners in the art of programming it served my own agenda of studying best ways to introduce observability in a microservice/serverless architecture quite well. So I managed to gather the following intelligence on thundra.io, the software product:

• Pros:

  • Support for: GoLang, NodeJs, Python

    • Supports a couple more

  • Allows integration through a serverless layer

    as a fallback there is also support for integration through a callback. note: with the layer, the runtime has to be set to provided as it’s manipulated by the serverless layer (thundra’s)

  • Comprehensive Dashboard

    • contains error filtering
    • contains coldstart status
    • shows the number of tries between services ( e.g. lambda → sns )

  • Can Target Invocation Tracing based on tags https://github.com/thundra-io/serverless-blog-site-workshop/blob/master/blogApi.js

  • Comprehensive Documentation

    https://docs.thundra.io/docs/getting-started-with-thundra-new

  • Integrates with HoneyComb

    https://docs.thundra.io/docs/thundras-honeycomb-integration

  • Can Obfuscate sensible data

    • through applying custom masks
    • by turning off response | request as needed

• Cons:

  • apparently only supports integration through serverless (hence the list of supported languages)
  • only works on aws
  • alerts are still a WIP (or alpha phase)
  • no public backlog of features/customer interaction
  • does not yet support grpc

• Main competitors / alternatives:

  • Lumigo - early days
  • HoneyComb
  • Zipkin

Keynote - Bringing WebAssembly outside the Web with WASI

In summary, super presenters Lin Clark and Till Schneidereit introduced myself in the world of WASI and helped shape ideas around a subject that although it’s far from being mainstream at the moment, seems to be in a prime position to help front end support from other programming languages than the mainstream javascript.

Lin had the best imagery:

She illustrated how files permissions are defined at the moment

And the new needs of security

while introducing WASI as a solution at the core of new front-end projects

Keynote - Serverless Attack Vectors

Teri helped raise awareness of security needs for production setups

In particular for aws lambdas

while also going through a series of relevant stories and examples that were recently in the media. As a result I started looking for a solution, but I did not get further away than this snyk blog post , with snyk showing a way of testing security pre-deployment.

Break

Here are my notes from all 3 breaks, it would be mean to say that the breaks were the most meaningful, but it wouldn’t be too far from it. Most of the event was static, while the 3 breaks were really dynamic with really good, well purposed booths.

Putting aside that I managed to nick some memorabilia from the kind people around the booths

I’ve also learned about

• Stardard Library and how good a replacement they are for IATT or Zapier; really keen to show this off at Flyt
• AWS Serverless Airline Booking, a superb tutorial series for building serverless architecture to serve a complete web app, their video series are an amazing fit for what this whole conference represents for myself and my future short term goals.
• OpenWhisk is too simple to be ignored, I will give this technology a chance in the near future

Lightning Talk

Machine Learning on Mobile and Serverless

A serverless app deployed using GCP

Speaker: Alexandra Abbas @alexandraabbas works at DataTonic as Data Scientist

• Keynotes

  • image classification model that can run on mobile ios and android

  • core ML does not perform too well on android

  • multiple challenges on mobile (device specific)

  • TFLite can execute models on mobiles with reduced functionality from TensorFlow

  • the model need to be retrained (manually -> on demand) regularly and deployed

Pipeline Def: upload (gcs) -> (data preparation -> Machine Learning) Cloud Composer

TFRRecords -> Train/Eval Sets -> Data Augumentation ->>

Apache Beam for Batch Data processing -> Dataflow (Python)

• Impressions

  • using gcp seems to be a lot more about dealing with the problems than the platform
  • you have to understand what components you can use in order to be effective

Discovering the microbiological world with Apache OpenWhisk and Rust

A serverless app deployed using OpenWhisk

Speaker: Robert Diaz @rdiaz82

• Keynotes

  • Story of Peter: the Quality Manager

  • provide an image as input, sequence of filters

  • openwhisk is based on docker -> easy to develop locally

  • internally: simple openwhisk architecture based on kafka as an event distributer

  • implemented using RUST

  • hold notes of improvements and future lines

  • try and make the platform generic and reusable

• Impressions

  • how wonderful it is to explain functionality with a story related to one use case

Serverless testing, the required adaptation to our testing methodologies

Speaker: lumigo.io

• Keynotes
  • Load Testing is a prime time for intoducing trace monitoring before production
  • serverless is a really effective technology for load testing

shop.LEGO.com

This lighting talk made me curious and sought further to discover lego’s journey about moving into serverless from the monolith. Here are my 3 vetted resources on lego:

• https://danilop.s3.amazonaws.com/Presentations/2019/ServerlessJourney+-+HEL+Revised.pdf
• https://twitter.com/sheenbrisals is so so cool
• https://www.youtube.com/channel/UCYzAnR_SebAmLRkKIbK_YoQ - after presenting to a couple serverless days events, I am sure one of lego talks will make it here (on the serverless days official youtube channel)

How AWS Builds Serverless Services Using Serverless

The Event Bridge announcement, stole the show

I wasn’t following the news coming out of AWS at the time, so this particular piece of news devised a new way of doing things in areas I was previously hacking around, I am hoping to match the Event Bridge with the video series discovered on the breaks (I bet they’re using Cloudwatch events).

A few other amazingly useful resources:

• https://www.trek10.com/blog/amazon-eventbridge/
• https://www.serverlesschats.com/5

My Key Take Aways from the event

• Trace Stacks are amazing for monitoring/debugging production setups, such are the ones introduced by thundra.io and lumigo.io
• I need to replace watching a tv show with these video series
• Replace my hacked used of Cloud Watch events with Event Bridge
• comments can be helpful