AWS in 2018: past is prolog
One of the publications to which I am a regular contributor recently asked me to jot down some predictions for AWS in the coming year. While I understand the appeal of this genre, I’m loathe to label something a prediction, with all the Solomonic connotations, since I feel that our inherent cognitive biases make it impossible to decouple coldly rational forecasts from hoped-for outcomes. Thus, I share this list with the caveat that they are an indistinguishable mix of things I think will happen and things I’d like to happen.
AWS is both intensely innovative and brutally competitive, with a customer centricity that means new products and services are often a reaction to customer requests and criticisms of existing offerings. While it can be counted upon to deliver a few surprises, many if not most of its new services will be enhancements to or evolutions of things that have come before. It also makes each fall’s re:Invent announcements a useful roadmap for where AWS is heading in the coming year. The problem predicting AWS behavior is that the company has no limits to the areas in which it competes. Much like its parent, AWS wants to win as much of your spending as possible. That said, there are a few areas I expect AWS to aggressively pursue in 2018, including:
- AI and machine learning: The three domestic mega-clouds are in an arms race for AI superiority with each introducing new applications (like speech recognition) and accelerated platforms (like GPU instances) at a furious pace in 2017. Expect this to continue, but with an emphasis from AWS on simplifying the development of custom applications using AI services. SageMaker is a good first step, but I would also expect to see packaged services like Amazon ML for other development frameworks like Caffe, MXNet and TensorFlow.
- IoT device and data management and backend applications: AWS introduced or enhanced 8 IoT services at re:Invent and I would expect an even greater number of improvements in 2018 as the cloud becomes the preferred backend for IoT applications and data. Look for additional work to improve support for native AWS services on embedded devices and its position on the network edge via new partnerships with device manufacturers and versions of Greengrass targeted to different use cases.
- ’Lightweight’ application infrastructure including containers and serverless functions: AWS pioneered serverless cloud services with Lambda, but Google and especially Azure have bypassed in key areas. I expect AWS to introduce an publish/subscribe bus for event management similar to Azure Event Grid and possibly enhance SWF (its workflow service) to make it more a low/no-code environment for application integration similar to Azure Logic Apps. Also look for Fargate (managed container infrastructure) to go GA with enhancements making it easier to migrate from Fargate to managed Kubernetes clusters (EKS).
- Enterprise integration and migration: VMware on AWS was the big news last year with a new Migration Service announced at re:Invent. While VMware’s service supports NSX virtual networks on its AWS infrastructure, this should be extended to make it easy to use NSX policies on native AWS VPCs outside the VMware SDDC. Although VMware’s NSX Cloud provides such network control via a third-party SaaS, it would be nice to see an integrated offering via the AWS marketplace that uses the native management console. Also look for the VMware partnership to expand to Pivotal with a service similar to the Pivotal-Google PKS that tightly integrates on-premise and AWS Cloud Foundry environments.
Also, lost amidst the current Bitcoin mania is the fact that blockchain is becoming a mainstream enterprise technology. AWS has been noticeably silent with only perfunctory support for partner solutions via AMIs. Look for AWS to introduce a managed blockchain service similar to the IBM Blockchain platform or Ethereum service on Azure.
Finally, expect AWS to maintain a steady pace of service updates and geographic expansion and to respond to price pressure from competitors, but not instigate a price war nor seek IaaS price leadership.