AWS Summit NYC Showcases API-Creation, Device-Testing Services for Developers

By Ellen Muraskin

Dr. Werner Vogels, CTO of Amazon.com, front-loaded his keynote address at the Amazon Web Services Summit in New York, July 9, with impressive statistics: 

Vogels also used the Summit to announce:  

Amazon Device Farm, an automated cloud testing service that help developers try out their applications on a full gamut of Android and FireOS smartphones and tablets at once. Reports and logs pinpoint places in code where functions fail, and on which devices.

API Gateway, a fully managed service  that  allows developers to create, publish, maintain, monitor, secure, and potentially monetize REST APIs more quickly, at any scale. These APIs make data, business logic or functionality available from “back end” services, such as code running on AWS Lambda or Amazon Elastic Compute Cloud. The gateway handles version control, access control and throttling, among other tasks.

Other recently released or soon-to-be released services, including AWS Machine Learning Services, got mentioned at the keynote. AWS ML walks developers through building a typical data model, gathering data, drawing inferences and making predictions; Vogels’ suggested use case was detecting fraudulent transactions.

He mentioned an AWS service for cloud cost management, with budgeting, forecasting and limit notification. This is a niche that was well represented by exhibitors on the show’s expo floor.

Vogels also tried to get his audience past thinking about servers; even virtualized ones: “Infrastructure is code,” he said. “Everything – databases, servers, load balancers – is software and has an API.” Then he brought on Keith Homewood, senior application architect at Nordstrom,, to describe how with AWS’ Route53 cloud-based DNS and CodeDeploy to automate code deployment, four people in their Team Cloud Center of Excellence support 50 application teams. 

The high-end fashion retailer is moving “all-in” to AWS, after moving its online “Recommendation” app, which suggests purchases based on personal profiles. Using AWS Kinesis data streaming and Lambda execution, the company reduced …

… 15 to 20 minutes of processing time into seconds, he said, and achieved a two-order-of-magnitude cost savings.  

Lambda is one of the latest in AWS microservices. “Forget about servers, instances, and containers,” said Vogels. “It’s too hard to estimate resources. Ideally, you don’t want to think about scaling. You just want to drop code somewhere.”

AWS Lambda allows users to write a bit of Java or JavaScript code and associate it with an event trigger, on automatically provisioned compute resources.  Using Lambda, the automated-bidding digital advertising company AdRoll makes bids on digital advertising space in under 100 milliseconds, while readers browse.

Cordell Schachter, CTO of the New York City Department of Transportation, spoke about his AWS-resident visionzeroview.nyc application, which gives citizens a constantly updated map of where accidents have happened most, in an effort to get drivers to obey newer, slower, life-saving speed limits. Another app, IRide NYC, runs on a hybrid on-prem and AWS infrastructure and gives users estimated arrival times by bus, train and foot.

To make the case for security being better in the cloud than it can be on premise, AWS brought onstage Mackenzie Kosut, CTO of the Oscar health insurance startup, who described his team’s DevOps achievement, building his company’s HIPAA-compliant, core systems with AWS in Linux in three months, producing 125 product changes per day.

Lower cost and eliminated capex is the “sizzle” that has attracted companies to ditch their data centers in favor of cloud, Vogels said. The lowered barrier to IT entry may also be the main factor for startups. But for most, the “steak” is agility, or speed and low cost of change.

With the ability to change on a dime – to stage compute environments and strike them at will – he stressed that companies have the ability to fail without penalty. To tweak resources among different instance types, and to experiment.