To paraphrase Douglas Adams, Re:Invent is big. Really big. There are more than 60,000 attendees this year and of those, 9,000 were in attendance at Andy Jassy's keynote speech yesterday. Keynote speeches like this are some of the biggest events in technology and this year the focus was on transformation.
To start, Jassy focused on what it is that companies require in order to successfully transform. This included senior leadership conviction and alignment, as well as leaders who drive aggressive goals to move toward production rapidly. But this keynote wasn’t really about the companies that AWS were helping transform, it was about AWS's transformation toward becoming a Machine Learning company, with a sideline in Cloud. Let’s look at some examples.
The SageMaker product set has been massively expanded, presented at such a pace that it was hard to keep up. First, there’s SageMaker Notebooks, to greatly simplify the use of Notebooks for one-click delivery. Then there’s SageMaker Experiments, to automatically capture, organise and search every step for model creation, training and tuning. SageMaker Autopilot delivers automatic training with no loss of control or visibility. The SageMaker Model Monitor automatically detects concept drift in models, while SageMaker Debugger helps developers improve the accuracy of models. And, probably most impressive, was the SageMaker Studio, a Cloud IDE for Machine Learning.
Frankly, this is an astonishing range of new capabilities, all either generally available or soon to be released. It's taking the ML game straight to the competition and strengthening one area where AWS' competitors had an edge on them.
New Machine Learning (ML) services
Jasser announced a whole new range of services using ML to deliver new capabilities. There's Amazon Kendra, a brand-new search function, which uses ML to search enterprise data with far better results than existing technologies. There’s Contact Lens for Connect, bringing ML-powered analytics to the contact centre, including sentiment analysis and the ability to review transcripts in near real-time. CodeGuru reviews code from GitHub and CodeCommit to identify issues, such as concurrency issues and input validation, while making remediation suggestions. And for Financial Services, there's Fraud Detector, an ML-driven real-time fraud management service.
Individually, these would be big news. As a whole, it’s a sign that AWS are refusing to rest on their laurels. After all, it doesn’t matter if you have 47% of the IaaS market when your competitors are catching you, undercutting you at every turn – you have to keep driving forward, understanding your clients and using a relentless delivery focus.
Compute and Networking
Speaking of IaaS, there were useful insight into some of the more traditional services. From a compute and networking perspective, it’s clear that the Nitro card has been upgraded, delivering major improvements in High Performance Computing (HPC) as well as the aforementioned ML products. This was covered in more detail by Peter DeSantis in the previous evening’s keynote. The success of the ARM-based CPU’s – which apparently came as something of a surprise to AWS – has driven the development of three new Graviton2 processors, delivering 40% price/performance advantage over the equivalent x86 chips. And then there’s the new Inf1 instances, specifically for rapid ML Inference.
There was an announcement on Outposts, Local Zones and VMWare, but nothing to directly compete with GCP’s Anthos, which was an interesting omission. Fargate saw a major bump with the ability to run Kubernetes, which will win many fans.
Also, AWS have seen Linux consumption grow 4 times faster than Windows, possibly driven by licensing constraints, which was mentioned with a certain degree of amusement.
On the database side, Aurora has become “by far” the most rapidly growing AWS service with a huge number of customers, many using it to migrate away from more traditional RDS with their cumbersome licensing models. There’s also new Managed Cassandra Service.
Finally, S3 Access Points looks like it will deliver a solution to the challenges of managing access to S3 buckets at scale. Redshift will now be able to manage compute and storage separately, with AQUA delivering major performance improvements by shifting the query handling to the storage layer.
In a nutshell
The last few months have seen both GCP and Azure make major advances toward delivering the huge range of services AWS provides, and there was a risk that AWS wouldn’t be able to make the step forward it needed to. It’s safe to say that, with the announcements around Machine Learning, AWS is taking steps to leap ahead of the competition once again.
Of course, the proof will come once large numbers of people start to use the services and determine what works well, and what doesn’t, and there’s no doubt that over the next year GCP and Azure will make further huge strides toward AWS.
In his closing statements, Andy Jassy warned companies that “if you just dip your toes in, you’ll fall behind.” Maybe that message was as much for AWS themselves as it was for their customers.