AWS’s big annual show in Las Vegas was predictably a torrent of launches. This year the headlines clustered around machine learning, databases, hybrid cloud and serverless — and those categories are exactly where cloud vendors are competing hardest. For ordinary users of web and mobile apps the takeaway is simple: apps are about to get smarter, faster and more flexible in where and how they run.
The announcements that mattered
Amazon highlighted expanded tooling for machine learning, notably additions to SageMaker that make it easier for developers to train and deploy models. It also unveiled its own custom inference silicon — announced as a way to accelerate the step where models are run in production — and further pushes into tooling that helps put machine learning into real apps.
On the infrastructure side, AWS announced Outposts, a family of rack-mounted hardware that brings AWS’s services and APIs into a customer’s own data centre. That is a clear signal: Amazon wants to offer the same cloud experience where customers need to keep data on-premises or minimise latency.
Databases also featured heavily. AWS refreshed and expanded its database options and tooling — everything from managed relational engines to fast key-value stores — emphasising performance, replication and easier migration paths for organisations that are moving workloads into the cloud.
Serverless computing, led by Lambda and related managed services, received further attention too. The theme across serverless announcements was about making it easier to run code without provisioning servers, connect it to other services, and scale it automatically.
Taken together, those are the headline areas: SageMaker and machine-learning tooling, custom inference hardware, Outposts for hybrid cloud, database enhancements, and more serverless features.
What this means for the apps you use
Smarter features: As machine-learning tooling becomes easier and cheaper to use, expect more apps to add features like personalised recommendations, smarter search, automatic photo or message tagging, and other predictions that run in the background. The key point is that building, testing and shipping models is getting less specialist, so features that once required big teams can turn up in smaller apps.
Faster, more responsive services: Custom inference hardware and broader deployment choices are aimed at lowering the time it takes to produce a prediction. For users that should mean snappier responses from features that rely on machine learning — think voice assistants, image analysis, or fraud detection happening in real time.
Data stays where it needs to: Outposts makes it practical for companies to use AWS tooling while keeping data on-premises or close to users. That matters for apps whose latency or regulatory requirements can’t tolerate public-cloud-only setups — they can run against the same APIs but in their own space.
Less ops overhead, more features: Serverless improvements let developers add functionality without managing servers, so updates and experiments can ship faster. For end users that usually translates into quicker feature rollouts and fewer service interruptions caused by capacity mistakes.
More choices and consolidation: The database announcements underline that cloud vendors want to own more of the stack. For consumers that’s mostly invisible, but it matters for how reliably and quickly data-driven features work.
Bottom line: re:Invent 2018 was another clear signal that cloud providers are competing on machine learning, hybrid flexibility and serverless convenience. You won’t need to worry about the technical plumbing, but you should expect apps to become smarter, faster and more varied in where they run.
This article has been restored to the What's New On The Net archive as part of the site's relaunch.