re:Invent 2017 : Four Highlights

By Sam Bashton – Head of Public Cloud, Claranet

Last week was the annual Amazon Web Services re:Invent conference in Las Vegas, where over 40,000 geeks from all over the world gathered to hear the latest news and major product releases (and a few jokes targeted at Oracle’s Larry Ellison). With so many feature enhancements and product releases in an average week at AWS, there’s always plenty to talk about. This year’s conference didn’t disappoint either, with an avalanche of great new technology announced.

I’ll pick just four announcements that I think really show where AWS are heading. In my opinion, although Google and Azure are fighting hard to catch up, it’s hard to see how they can get anywhere close when AWS are playing the strategic game so well. From a technical point of view, many of the new features AWS released match those Google announced at their Next conference last March, but the breadth of what Amazon offers are unrivalled. AWS has an answer for basically every workload, including some (IoT for example) where they have stolen a significant march on their competitors.

EKS

EKS – or ‘Amazon Elastic Container Service for Kubernetes’, to give it its full name – was a highly anticipated release which I believe really shows how differently AWS behaves to the rest of the IT industry.

Kubernetes is an open source container orchestrator initially created by Google, and offered as a managed service by them since 2014. Initially, Amazon tried to compete with their own open source project, Blox, but they have clearly been listening to what customers are asking for (“Make Kubernetes on AWS easy!”)

The ability and willingness to adapt, instead of stubbornly doubling down on their own competing product, provides real insight into why AWS will continue to dominate the Cloud landscape for the foreseeable future.

Neptune

Graph databases aren’t at the forefront of IT media consciousness in the way machine learning is, but they can offer many of the same capabilities.

The “People that bought this also bought…” alert is often trotted out as an example of machine learning, but a graph database can provide this data without needing to be trained on large quantities of data. Up until now, running a clustered graph database in production such as Neo4j was possible but had a sizeable management overhead. Neptune brings the ease of RDS to graph databases, with the same underlying functionality of Amazon Aurora storage. In addition, data is replicated across multiple data centres and read replicas are provided, but in a fully managed service.

I expect Neptune to bring a new interest to this under-appreciated class of databases. When combined with machine learning enabled from Sagemaker (below), Neptune provides a way for businesses to easily build tools that were previously out of reach for anyone but the tech giants.

SageMaker

Machine learning is complex, and Amazon SageMaker doesn’t change this fact; it won’t be putting any data scientists out of work, but it will make them more productive. What it does is take away much of the ‘undifferentiated heavy lifting’, making it easy to spin up clusters and import, explore and visualise training data.

In a similar vein to the EKS announcement, it’s interesting to note that Tensorflow, another Google open source project, has equal billing with MXNet, the open source project behind which Amazon has previously put most of their energy.

Deeplens

I’m constantly surprised by just how much of a lead Amazon’s competitors have let them steal in the IoT space, particularly given the fact that the people Amazon are up against are market leaders in mobile and desktop.

With AWS IoT and Greengrass, announcements from re:Invent in previous years, AWS have created a great environment to build IoT devices. Deeplens is effectively a hobbyist device, which demonstrates just how effective these can be when combined with Sagemaker. Essentially a small Linux device with a built-in camera and a powerful graphics processor, Deeplens has been built to let people play with computer vision machine learning. At re:Invent, Amazon got Deeplens to detect people and hot-dogs in just a couple of hours, and I think this is an amazing tool to let people easily and cheaply prototype new features.

Find out more.

Leave a Reply