logo
Menu
My Day at AWS Summit London

My Day at AWS Summit London

A summary of the talks I attended at AWS Summit London and what I took away from them

Published Apr 29, 2024
I was fortunate enough to attend my lucky first in London on the 24th of April. This was a 1-day free event in the Excel Center. I was blown away by the scale and quality of this event, I was fortunate enough to attend AWS re:Invent in 2022. This felt like a 1-day version of that (minus the maze of casinos!)
In this blog, I'll share my favourite takeaways from the talks I attended, so you can get a flavour of what happens at an AWS Summit.

Keynote

AI and sustainability were the main focus of the keynote, here's what I took from it:
Moderna sequenced the COVID-19 vaccine in just 2 days using AWS, with the first vaccine being released 25 days later
This completely blew me away, a few years ago the race to develop a vaccine was world news. This is a great example of how leveraging the scalable computing power of the cloud can have life-saving impacts.
AWS is 5 years ahead of their goal to be net-carbon-zero. The original goal was 2045, but now it looks more like 2040.
Moving an on-premise workload to AWS reduces carbon emissions by up to 80%
Obviously, with data centres consuming an ever-growing percentage of the world's energy supply, AWS must work towards lowering the climate impact of their operations. It is good to see that they are progressing better than initially expected. You can read more about AWS Sustainability here.
Zilch increased development speed of AI models for fraud prevention, credit underwriting, customer service and buyer intent prediction by leveraging AWS Bedrock and AWS SageMaker
Zilch is a direct-to-consumer ad-subsidised payments network, focused on eliminating the cost of consumer credit. They are using these models to ensure they can protect themselves from unnecessary risk and to make underwriting decisions for credit applications. I found this extremely interesting as a lot of these use cases apply to many industries (including the insurance industry that I work in).
You can read more about Bedrock and SageMaker at these links.

Keynote session
Keynote Session

Experience the future of development with next-generation developer tools

This talk focused heavily on AWS CodeCatalyst which is a "unified software development service" that allows developers to speed up development by leveraging AI capabilities and by quickly configuring development environments.
My main takeaways from this were:
CodeCataylyst can manage the entire development lifecycle, from project management, local development and build and deployment steps.
While I had heard of CodeCatalyst, it seems like it's gotten several more features since its initial launch. You can find out more about CodeCatalyst here. The project management tab is a nice addition, with integrations to show the different stages that task is at in the development lifecycle.
Leverages your IDE as a thin-client, reducing resource usage on your local machine. You can specify what scripts to run when you set up your remote development workspace, to install dependencies or any other tasks you want to run.
You can spin up a remote workspace easily, configuring CPU and memory as needed. Running locally is easy as the port forwarding is already configured. This could make development more easily accessible for those with lower-specification machines. The scripts you run when starting these workspaces make the developer experience quick and painless. Forget remembering which commands you need to run so you can focus on what matters most.

Building Obseravility to Increase Resiliency

This talk focused on building best practices for engineering excellence around observability.
My main takeaways from this were:
Consider Dimensionality - Show me [metric] per [dimension]. Metrics could be latency, requests, error counts. Dimension could be consumer, instance ID, availabilty zone.
Image from the presentation giving examples of measurements and attributes that could be applied
Examples of how to consider dimensionality
This was one of my favourite talks of the entire day. For example, if 1/3 of instances became unhealthy and you only looked at the overall picture (did not apply dimensionality) it might go undetected.
If you have 2 consumers, and 1 of them is 99% of your traffic is healthy, you might not notice that the consumer who makes up 1% of your traffic is having issues. Look at what the impacts are for each customer and not only the bigger picture to add dimensionality to your monitoring to improve the customer experience.
They also spoke of the importance of engineers owning applications end-to-end and being responsible for supporting them 24/7 - it gives a vested interest in building stable and reliable applications if you know you could get woken up to address an issue in the middle of the night.
Metrics make you aware, but logs tell the story
This was another great quote from this talk - you are typically alerted to issues through metrics but you will often go to the logs for the next steps once you know an issue is occurring.
Ensure your logging is useful and meaningful and you are not logging irrelvant information
Each log ultimately has a cost associated with it, but over-logging can also increase the noise when you are attempting to troubleshoot issues.
I definitely took a lot from this talk, including the observability guild JustEat has, the use of composite alarms to reduce alarm fatigue and how to build observability that is valuable to your engineering teams and customers.
Step up your game - simplifying complex workloads with AWS Step functions
Hamilton Robson was building a management system for the construction sector, which involved validating a few documents such as a CSR card, National Insurance number and passport. The application was originally lambda orchestrating to other lambda functions synchronously.
The customer returned with new requirements to make some of these checks optionally, while they initially added additional branching it became complex.
Lambda should only be used to transform and not only to transport
This is a great quote to remember and is part of the reason they decided to re-write the application as a step function.
Hacking in a change the customer wants
Hacking in a change the customer wants
By rewriting the application as a step function they were able to support parallel processing of each of the required validations, with the responses collated and validated at the end.
The step function workflow can be visualised using the Visual Studio code extension, making it easier to validate your application and what you expect without needing to deploy.
Monzo's Journey to Amazon EKS with Karpenter and EC2 Spot
This talk started with someone from AWS talking about the power of EC2. Nitro is the underlying management software AWS uses to ensure optimal resource usage of AWS compute so that they can run the least amount of servers to provide for their customers.
When Monzo was expanding they went from 60k new customers per month to 200k new customers per month, so traffic increased massively and they needed to scale their infrastructure to handle this.
They previously managed their own EKS clusters, but they switched to managed EKS to focus on more value-adding work for their customers.
They also adopted Karpenter to manage the auto-scaling of their services. They used savings plans to deal with predictable traffic patterns as they are primarily a UK bank.
Monzo's traffic pattern / Savings Plans
Monzo's traffic pattern / Savings Plans

Closing Thoughts

I enjoyed my time at the summit, it was a busy day. Talks are first-come-first-served basis, some of the smaller talk areas such as the Apollo Partner Theater and Community Lounge filled up quickly so I missed a few talks I wanted to attend.
There were lots of fun activities to participate in in the expo-hall such as an F1 experience where you could change the tyres on an F1 car and drive a racing simulator. There was also a penalty kick football challenge that used tracking software to measure the speed and positioning of the ball when you kicked it. These had long queues all day, I tried twice for the F1 stand but decided to attend more talks instead.
I’d recommend planning your day with the AWS Events app and making sure you know what path to take to get to your next talk.
I had a great time at the AWS Community social event after the Summit, I spent a lot of time talking with the founders of LocalStack, quizzing them about all things from starting your own business to growing to 40+ employees worldwide.
AWS Community Photo
AWS Community Photo
If you haven’t attended an AWS Summit before, I’d recommend you check out this link to see what one is near you.
Let me know if you attended the summit and what you thought of the day in the comments!
 

2 Comments