Yesterday, Dr. Werner Vogels, CTO of Amazon.com, kicked off the morning keynote by interacting with Amazon Alexa. Alexa asked about his worst day at Amazon, which he said was December 12, 2004, the cutoff date to get packages out for Christmas for their Super Saving Shipping promotion. To put it simply, the database was not equipped to handle the volume of orders and having even a single point of failure was devastating. Vogels shared the story of his worst day as a call to action for users to know the architecture behind the services they are using and how to build in a way that minimizes the impact of a failure.
The next generation of database systems
Following the story of his worst day, Vogels dived into the architecture of databases by first introducing sharding as a way to reduce your ‘blast radius’ from a failure, and how AWS began using cell-based architecture so they have total control in case of failure. Amazon Aurora, referred to as the foundation for innovation, was highlighted for its scaled-out distributed architecture and its ability to re-replicate and return the database to its healthy state just 10 seconds after any failure.
Vogels then introduced the four pillars of Amazon DynamoDB—performance at scale, full-managed, comprehensive security, and built for modern applications. He set the stage for a comparison of the Oracle schema and the DynamoDB schema and illustrated how to handle spikes and bursts in capacity with automatic re-sharding. The next generation of databases topic was concluded with a highlight of the AWS Database Migration Service which lets you migrate with minimal downtime.
The AWS culture of durability
The Vice President of Amazon S3 and Glacier, Mai-Lan Tomsen Bukovec, came to the stage to share the architecture of Amazon S3. When S3 was launched 13 years ago it had 8 microservices, it now has more than 235 distributed microservices. She stated that having a culture of durability is of the utmost importance and for every new feature AWS builds they do a durability review. This is similar to a security threat model reviewing for loss of data. Common tasks associated with the review include a static analysis, checksums and proofs, durability checks, and operational safeguards. Tomsen Bukovec described how S3 and Glacier are designed for 11 9’s and for AWS the durability model is all about the actual time to repair. She then wrapped up her segment by emphasizing there is an entire data center dedicated to S3 durability.
Time for the best day
Earlier we touched on Vogels worst day, but when he returned to the stage Alexa asked him what his best day was. He said his best day was November 1, 2018, which is the day they switched one of the world’s largest Oracle data warehouses and moved it over to Amazon Redshift (if you haven’t noticed yet, the AWS keynotes like to take shots at Oracle). Vogels noted that 87% of Redshift customers have no wait times and in an effort to increase this number he announced the preview of Amazon Redshift concurrency scaling which provides consistently fast performance so you never have to wait or see a charge for auto-scale resources. Vogels then brought Ethan Kaplan, CPO at Fender Digital, Fender Musical Instruments, to the stage to share how Fender is all in on AWS and have built applications including Fender Tune and Fender Play to help guitar players learn and play. Before exiting the stage Kaplan presented to Vogels a Fender guitar customized with the AWS logo.
Users are flocking to serverless
Now for the topic the audience had been waiting for: serverless. Vogels stated, “Serverless allows you to build increasingly powerful applications faster which is why customers are flocking to serverless.” Holly Mesrobian the Director of Engineering for AWS Lambda was invited to the stage to share how AWS Lambda is available in every AWS region and processes trillions of requests every month. She then dove into some of the key components such as Workers to provision a secure environment for customer code execution and Firecracker which was announced during Monday Night Live. Firecracker provides secure and fast microVMs for serverless computing without negatively impacting performance. Mesrobian emphasized that AWS is making deep infrastructure investments to support serverless.
Vogels then returned to the stage to discuss how everyone wants to focus on business logic and 95% of AWS features and services are being built based on customer feedback. This introduction marked the beginning of numerous serverless announcements across IDEs, languages, programming models, and workflows. First, Vogels announced the AWS Toolkits for popular IDEs which include PyCharm, IntelliJ, and VS Code. Then he announced the general availability of Ruby support for Lambda. Next, he announced Custom Runtimes which gives you a platform to bring your own favorite language and is powered by a new Runtime API. This means there is now no limitation to what language you can use for serverless development. The general availability of Lambda Layers came next which extends the Lambda execution environment with any binaries, dependencies, or runtimes.
Vogels went on to introduce other exciting enhancements including:
- The Nested Applications Using Serverless Application Repository which lets you compose application architectures from reusable building blocks and share applications with teams or publicly to improve agility.
- Building on the release of Step Functions from a few years ago, Vogels announced the general availability of Step Functions service integrations which let you connect and coordinate AWS services together without writing code, including AWS Batch, Amazon ECS, AWS Fargate, AWS Glue, Amazon DynamoDB, Amazon SNS, Amazon SQS, and Amazon SageMaker.
- Next was WebSocket support for API Gateway which lets you build real-time two-way communication applications. This allows you to transition from a traditional compute environment to the serverless world without having to change your APIs.
- Vogels then announced ALB Support for Lambda which enables you to integrate Lambda functions into existing web architectures, removing more barriers to entry.
- The final announcement on the serverless topic was designed to alleviate challenges with deploying Apache Kafka, thus the Amazon Managed Streaming for Kafka is now in preview.
Are you well architected?
Coming full circle with the architecture theme for this keynote, Vogels started to close out by asking “Are you well architected?” He then welcomed National Australia Bank to the stage to share their story about how they are building on AWS. Vogels then highlighted the five pillars of the Well-Architected Framework which includes operational excellence, security, reliability, performance efficiency, and cost optimization. Then he discussed the current review process and how there is a new competency program for APN Partners to be able to perform the Well-Architected reviews.
In closing, Vogels made his final announcement of the day which was the availability of the AWS Well-Architected Tool which enables you to measure and validate your own architecture using AWS best practices and gain deep insights across your technology portfolio.