AWS kicks off re:Invent with compute, networking, and data launches

For anyone expecting the new Amazon Web Services’ (AWS) CEO Adam Selipsky to beat a new path for the cloud behemoth, Tuesday’s keynote address at AWS’s annual user conference would have come as a disappointment.

The former Tableau CEO -- who returned to AWS earlier this year when Andy Jassy ascended to the Amazon CEO throne -- was never going to be charged with setting a drastic new direction for the cloud behemoth, which maintains a healthy double-digit market share lead over its rivals Microsoft Azure and Google Cloud.

However, a pair of pre-conference interviews with Bloomberg and the Wall Street Journal hinted at a simplified, industry-focused approach from the new CEO, who held up Amazon’s cloud contact centre product Amazon Connect as an ideal example of how AWS could simplify its offerings for less cloud-savvy customers in the future.

As Matt Asay wrote for InfoWorld ahead of the show, “It’s a safe bet to take Selipsky at his word and assume the company is going to try to tame its Byzantine service creep (more than 200 services and counting) for customers with vertical industry solutions.” While taming the Byzantine service creep may be forthcoming, it was not on display the first day of re:Invent.

Major AWS re:Invent announcements

Even though he spoke for an hour less than his boss typically used to, Selipsky took a leaf out of Jassy’s playbook by continuing to make a massive number of announcements during his re:Invent keynote, focusing on three key areas: compute, networking, and data, in that order.

This included new compute options, such as C7g, a new EC2 instance powered by Graviton3, the latest iteration of its Arm-based custom silicon portfolio. These instances promise to operate 25 per cent faster on average for compute workloads and better still for specialised cryptographic or machine learning workloads.

Another instance called Trn1 was launched in preview, which is the first to use the Trainium chip that was announced last year. Trn1 promises higher performance for compute-intensive machine learning training use cases, complete with 800 Gbps of network bandwidth capacity.

Selipsky also set his sights on a large cohort of customers who still run Cobol applications on mainframes through its new AWS Mainframe Modernisation program. “Many customers are trying to get off mainframes to get the agility and elasticity of the cloud,” he said.

Credit: AWS

Regardless of whether customers want to simply re-platform or totally refactor their existing mainframe applications into microservices, AWS says it can reduce the time it takes to migrate by up to two thirds by automatically recompiling code to run on a new “mainframe-compatible EC2 runtime” or automatically refactor the Cobol code to Java for you, Selipsky said.

As the developer and ex-AWS employee Tim Bray tweeted during the keynote: “Big enterprise customers with traditional apps are where it’s at. The world of cloud-native people/apps is invisible.”

On the data side, the big launch was SageMaker Canvas, an extension to the cloud vendor’s existing machine learning workbench that aims to enable non-developers to build machine learning models using a point-and-click interface. AWS also launched serverless options for the Redshift, Elastic MapReduce, Managed Streaming for Apache Kafka, and Kinesis analytics services all at once.

Not announced on stage, but potentially highly interesting for developers, was the launch of a new open source Kubernetes cluster autoscaling tool called Karpenter. 

Karpenter is aimed at helping developers move away from manual cluster right-sizing by providing “just-in-time compute resources to meet your application’s needs and will soon automatically optimise a cluster’s compute resource footprint to reduce costs and improve performance,” principal developer advocate Channy Yun wrote in a blog post.

What next for Selipsky and AWS?

There is of course the possibility that this year’s event, which landed during the middle of a global pandemic, came at the wrong time for Selipsky to make any major changes to a business that could realistically hit US$60 billion in revenue this year.

The new CEO did hint at his vision during his final comments, where he parroted Jassy in saying that AWS is “only getting started” and that there is still plenty of cloud market to be captured by making AWS “more powerful, more intelligent, and easier to use.” The last of those vows may just be his biggest test to come.