Key things to know about AWS re:Invent 2018 (part 1 of 2)

The team here have been watching the annoucements at the conference with a great deal of interest and we have noted some common themes:

Data, Machine Learning & AI

Hybrid, hybrid, hybrid

The push to expand the offerings in the data space make absolute sense, looking at data and data analytics, and borrowing from a Amazon phrase, we really are at Day 1 with data in most organisations. Cloud is a known quantity and really a key enabler for the foundations of innovation, and this has enabled a more sensible approach to how we can innovate on the data we already have in the organisation and enrich it with other data sources to blow the socks off the customer.

The aim here is to talk about the announcements at re:Invent itself, so here goes! I recommend you read through Jeff Barr's post, which can be found here.


AWS Inferentia

AWS have created dedicated hardware as we have pushed past the capabilities that general chip architecture can support in this space, and specialist hardware is the best way to support this ever expanding space, this specific capability is called AWS Inferentia.

Inferentia comes from the machine learning approach of inference, which is the act or process of deriving logical conclusions from premises known or assumed to be true.

Developers can reduce inference costs by up to 75% by attaching GPU-powered inference acceleration to Amazon EC2 and Amazon SageMaker instances.

AWS Marketplace for Machine Learning

When Sagemaker was launched it became the fastest growing services in AWS history, enabling engineers to have clear ability to draw down on, allowing easy access to creating new ways to better support their customers, below is a view of the marketplace showing that there are already 489 different items (364 of these in Sydney) available:

There are some interesting options for applied machine learning approaches here, AutoScalr is an AutoScaling service that applies machine learning to save operational costs on autoscaled applications by determining optimal use of spot instances spread across multiple spot markets and On-Demand instances to handle production workloads without requiring code changes. I would expect to see the number of options double if not triple in a very short period of time!

I have to mention AWS DeepRacer, this uses reinforcement learning (where the end state is known but there is little training data available to develop that outcome, using an interactive environment to learn over time). It includes a fully-configured cloud environment that you can use to train your Reinforcement Learning models. There will be a league for racing in over the next year, winners and top scorers will advance to the AWS DeepRacer 2019 Championship Cup at re:invent 2019. Very cool!!


AWS Outposts

There are companies that have requirements to host systems in their datacenters, whether that is due to the tyranny of distance (mining) or in the case of data compliance obligations (financial services). AWS has responded to this by creating the ability to have AWS Outpost allowing hardware in these locations in two variants:

A) VMware Cloud on AWS Outposts allows you to use the same VMware control plane and APIs you use to run your infrastructure. Further deepening the relationship with VMware (SPECULATION: Could VMware be bought by AWS), VMware have provided more information on capability here.

B) AWS native variant of AWS Outposts allows you to use the same exact APIs and control plane you use to run in the AWS cloud, but on-premises.

Both variants allow for the native running of AWS services, hardware will be shipped to a customer’s on-premises datacenter and connected to their power and networking infrastructure.

AWS Transit Gateway

In my experiences networking with the cloud is one of the most complex areas to both implement and maintain. The reason I mention this service under hybrid, is that with multiple VPCs and a hybrid solution with an on-premises environment, routing between VPCs and on-premises can get complicated. You have to set up VPC peering. 

This allows you to implement a hub and spoke model within AWS. You can connect all your VPCs and on-premises environment to the Transit Gateway. Routing is taken care of automatically. 


I have said it before and I will say it again, the ecosystem around AWS is phenomenal and the flexibility of using the platform just keeps growing. I wanted to call out another feature (although many more were announced), AWS Private Marketplace. if you want to know more click here. This allows administrators to choose from thousands of listings in AWS Marketplace to create a curated list of products their users can purchase.

The conference is not over yet, I know there is a lot more to come! Have fun, and happy to chat as always about any and all of these updates!

Re-posted from LinkedIn, here -


Recent Posts

See All
atlasplato blue.png