AWS Transformation Day BENELUX
Taking a business to the cloud often requires a transformation. Where should a business focus on? What tools should be used? What are the pitfalls of transitioning? I went to NBC in Nieuwegein for the Amazon Web Services Transformation Day BENELUX to find out.
A broad range of topics is covered in six different tracks ranging from deep dives in serverless development and big data architectural patterns to EU privacy compliance and business case development for AWS.
Being a developer and an advocate of innovation within Moddix, I dove deep and went for a nerd track selection covering the following topics:
11:20 — 12:00 Introduction to Amazon AI
Companies like Google, Apple, Microsoft and Amazon all agree on it: we are moving from mobile first to artificial intelligence (AI) first. Machine learning (ML) and deep learning (DL) will be incorporated in our day to day lives. It is our responsibility as creators and innovators to start building products that leverage these techniques and bring AI to the consumer.
Julien Simon tech evangelist at Amazon and Wojciech Soltysiak chief technology officer at CHAMP gave us an exciting look into some of Amazon’s AI services. Given the difficulty of developing your own ML models the use of these services enable developers to use AI technologies without having to roll their own.
Wojciech started off by showing their customer support product powered by Amazon Lex (for chatbot interaction) and Amazon Alexa — the (in)famous personal assistent. Although Alexa has seen a lot of attention when it comes to ordering products from Amazon’s own store, the customer support case sounds very compelling.
Since CHAMP is in the air cargo business, their primary use cases for customer support are:
- Creating shipments
- Retrieving shipment status information
Both cases were demo’d without a single error. Retrieving shipment status felt like a phone call with a well trained customer support employee and gave a rich description of some shipments set up prior to the talk. Creating a new shipment felt solid. Asking the right questions and providing the user with enough feedback/confirmation to be confident in the correctness of steps taken by Alexa.
Exciting tech, well suited for customer support, what’s up next?
Julien followed up with a demo of a small robot taking pictures of the audience. The raspberry pi driven robot started talking: “The picture shows three people, age between 32–43, 99% happy, 1% surprised, one man wearing glasses.” This all was realised with a 30 line python file calling a single Amazon API endpoint.
13:05 — 13:50 Big Data architectural patterns and best practices on AWS
Large projects often produce big data. Handling big data can be complex and a well designed big data architecture can reduce a lot of this complexity. Apart from the complexity involved, being able to be cost-conscious is something to keep track of at all times. Nobody want’s big data to turn in to big costs.
Franscesco Rinaudo, Senior Manager Solutions Architect at Amazon takes us through the “Big Data Evolution”. Where batch processing used to be the dominant strategy, big data systems evolved to stream processing and are now taking the step to ML based processing. Key takeaways of the talk include:
- Build decoupled systems
- Use the right tool for the job
- Leverage managed services
- Use log-centric design patterns
- Be cost-conscious
Next up Dimitar Nedev, Data Alchemist at PiCNIC Supermarkets shared his experience with “Scalable data analytics for Mobile Commerce”.
At PiCNIC they use explicit and implicit user feedback to help shape their storefront. Analysing this feedback helps them focus on the most used features and improve general customer experience. In the past two years they saw an increase in feedback events from only a few million to more than fifty million a month and increasing.
Plowing through these amounts of data requires a solid data processing strategy. Initially they used a collector and pre-processor running inside an Elastic Container Service (ECS ) cluster connected via Amazon’s stream collector and processor Kinesis. This allowed them to efficiently pipeline the data to S3 for further batch processing with the help of spot instances.
To further reduce cost however, they moved from using spot instances to AWS Lambda. This not only enabled them to reduce cost but also split streams for both realtime data analysis & rule based processing as well as data lake storage for later analysis using Amazon Athena.
13:55 — 14:40 Deep dive on serverless application development
Businesses need to adopt to changes in a growing pace. Rapidly deploying new services and being able to scale up and down based on demand are name of the game. Serverless application development is the latest of innovations and brings great promise:
- No servers to provision or manage
- Scales with usage
- Never pay for idle
- Availability and fault tolerance built in
Prakash Palanisamy, Solutions Architect at Amazon walks through a couple of use cases and dives deep into the deployment and management of AWS Lambda functions. Touching configuration management, build pipelines versioning and logging he finishes his talk with a small demo application showcasing the power of AWS Serverless Application Model (SAM), AWS CodeBuild and AWS CodePipeline.
15:15 — 16:00 Scaling up to your first 10 million users
Transformation and scaling go hand in hand. An organisation can profit from an architecture designed for scale while organisations failing to scale have difficulty adopting to changing demands.
Adrian Hornsby, Technical Evangelist at Amazon takes us on a journey from having an idea to having ten million users from across the globe highlighting the importance of choices made early on.
Are we building an enterprise application or a simple web app? Will it be used across different time zones? Is the content static or dynamic? From day one on, Adrian advises operational excellence:
- High quality, version controlled code
- Have a CI/CD pipeline
- Infrastructure as code
- Security at every layer
- Be cost conscious
- Test everything
- Disaster recovery (DR) procedure at hand
What follows is a series of infrastructure layouts driven by standard AWS components. Starting out with a single static web page delivered by S3 and ending with a well thought through decoupled infrastructure handling multi region, highly available environment capable of serving millions of users.
When it comes to scaling to ten million users the following key concepts should be noted: use managed services where possible; offload compute intensive tasks to queue workers; leverage serverless computing to be cost conscious while increasing decoupling.
Closing out the day are Frank van Boven & Roelof Reitsma, Team Lead & Senior Software Developer at Coolblue.
Their story starts at 1999, the inception of Coolblue. Followed by a fresh start in 2004 after struggling with their home-brew software roots, kickstarting their journey towards transformation to the cloud.
Creating a new project made sure they had the right foundation, but it was by no means ready to scale. Issues like disk space- and bandwidth usage or the lack of configuration management when they started buying their own hardware started to appear once their customer base grew. On top of it all, their business experienced peak usage during holidays which gave them a hard time provisioning enough machines to stay up and running.
Solving the peak hour scaling issue concludes their story. A combination of Puppet for configuration management, AWS CloudFormation for provisioning and Github allowed them to move their core architecture to AWS, going live today! 2017 will be the first year they surpass the 1 billion euro revenue mark. Transforming to AWS allows them to scale effortlessly into the future.
AWS Transformation day was a smooth experience. Combining talks with enough depth to learn a few things and a broad selection of topics to entertain both the true geeks as well as senior management I would recommend this event to all who missed this years edition.
