Cloud News Cloud Updates Migration

Highlights from AWS re:Invent 2017: Instance Types, Containers, and More

Well it’s over: another re:Invent is behind us. The biggest annual conference for Amazon Web Services attracted over 45K people this year, spanned four hotels, and was in a single word: overwhelming. I’ve been going to re:Invent each year and it’s been staggering to see the growth in attendance.

If you’ve never been to re:Invent, the structure is basically the same:

  • Day 1: Bootcamp and certification exams, as most people are just arriving.
  • Day 2: This day is partner focused, with MSPs, ISVs, SIs and VARs getting together for industry specific sessions. It includes a keynote address from Terry Wise, VP of the AWS Channel.
  • Day 3: Now the conference kicks into high gear with CEO Andy Jassy’s keynote.
  • Day 4: The always popular CTO Werner Vogels keynote (and his ever present T-shirt and sneakers). The evening of day four always ends with the big AWS hosted re:Play party, DJed by some cool new musical artist unknown to anyone over the age of 30.
  • Day 5: The tired and hungover who just can’t get enough of Amazon go to their final sessions before flying home.

The News from re:Invent

Amazon always packs the show with announcements, and these typically occur in the two keynotes. This year, Amazon had so many new services being released that they announced them in a few other venues. Although not a comprehensive list, here are some of the highlights.


Containers are a huge focus among cloud customers, and Amazon responded with several new services including AWS Fargate, to allow businesses to leverage containers without managing the orchestrator or cluster fabric, and Elastic Container Service (EKS), to provide managed Kubernetes. Since Kubernetes is a Google open source project, I’m sure folks in Mountain View were not entirely pleased to see EKS. If imitation is the best form of flattery, Amazon was really laying it on thick for Google this year.


Andy Jassy likes to poke the Redwood City bear, and almost always dedicates time to Ellison and his crew during the re:Invent keynote. Amazon is looking to topple the Oracle database monopoly, and you don’t need to look further than their announcements at re:Invent to see the scope of their investment. I was particularly interested in Aurora Serverless, which provides a full MySQL / Aurora database without having to worry about infrastructure. Their graph database Neptune was also cool, as was Dynamo Global Tables, which is directly competitive with Google Spanner.


New Instance Types

Amazon released M5 (general purpose compute), H1 (high disk I/O), and T2 Unlimited instance types, as well as a new Bare Metal offering. The latter is particularly interesting since it allows customers to provision hardware like a more traditional hosting company (e.g. Rackspace, IBM SoftLayer). They are clearly leaving no stone unturned in their goal to operate the world’s applications.


Say this in your best Jan Brady voice: Lambda, Lambda, Lambda. Amazon continued its serverless investment with a flurry of Lambda announcements, including increased memory capacity, concurrency control, AppSync, Serverless Application Repository and more.


Cloud9 was probably the most eye catching announcement of the conference. It is an Integrated Development Environment (IDE) that makes it easier for developers to create applications on AWS. The service is the result of an acquisition Amazon did earlier in the year. They demonstrated some cool features, including pair programming and real-time collaboration. (An interesting aside: Almost all the speakers in Werner’s keynote were developers under the age of 30, in contrast with previous years. I can guarantee that no previous re:Invent speaker would have had any clue who DJ Snake was.)

Sumerian was also kind of cool, allowing for the rapid development of virtual reality, augmented reality and 3D applications. I’m thinking this may help me achieve my goal of seeing the world with a Clarendon filter.

Machine Learning

If last year’s hot topic was serverless, this year’s was machine learning. They unleashed a torrent of announcements including services to do real-time translation (Translate), transcription (Transcribe), video recognition & analysis (Rekognition for Video), natural language processing (Comprehend), and more. They even announced DeepLens, which is a hardware image recognition device that you can put in your living room. Downside: it’s creepy. Upside: it’s less creepy than the equivalent Google product. The big question at the conference was which we would run out of first: new Amazon ML services or words in the English language to use as AWS service names.


The IoT announcements that were most interesting to me included IoT Device Defender and IoT Analytics. Andy Jassy called out IoT as one of the areas in which he feels partners can add high value to Amazon’s customers (the others were: databases, machine learning, and migration). With no sense of irony though, he also said (paraphrased): build cool stuff on IoT, but hey, we’re doing this too so hopefully it all works out for you.


Amazon GuardDuty is a new machine learning-driven threat detection service that monitors your infrastructure for anomalous traffic. I’m looking forward to checking this one out. This makes great sense and could be a high value service for all AWS customers. Notice the use of the words “machine learning”. Machine learning. Machine learning. Did you hear this is driven by machine learning? Yeah now you are feeling like you were at re:Invent.


Amazon has put forward a credible solution to the big data problem they unleashed with S3. Businesses quickly discover that managing a few million objects on S3 is hard, but managing billions is impossible. S3 has become an infinite storage locker that gets so cluttered no one bothers to remove anything any more. S3 Select and Glacier Select provide services to allow you to apply SQL queries to your objects.


Alexa for Business enables businesses to leverage Alexa for workplace productivity. With its six Alexas, my home is an NSA listening post right now…so why not bring Alexa to the office, too?


Parsing through all the information from re:Invent is no mean feat…especially when sessions are scattered across hotels, which means choosing between an overflowing taxi cab line or a bus ride that sometimes takes 75 minutes (here’s hoping Amazon handles the sprawl differently next year, considering how large it’s become).

Transit complaints aside, here are some of my initial reactions and top takeaways, based on the buzz at the show.


AWS wrote the functional spec for IaaS the industry has embraced over the last several years. They innovated and dominated on everything IaaS, and much of their revenue today is still driven by compute and storage. But there was an unmistakable shift in their focus to PaaS at this re:Invent. If you have heard me talk about Cloud 2.0, you can mark this week as the beginning of the true shift away from 1.0 (IaaS).

Cloud 2.0 is about leveraging the true power of cloud computing — platform services, dynamic optimization, arbitrage, serverless — to fundamentally change how you build, deploy and deliver applications. It allows you to harness cloud agility and innovation to drive radical business transformations. To support this shift, this is the first re:Invent where we heard pervasive use by Amazon of the words “managed” and “fully managed” related to service offerings. The future is all about abstracting infrastructure.


In the past there was a generally agreed upon rule that while at re:Invent, you act like there was no cloud other than AWS. It’s sort of like Fight Club: the first rule of re:Invent is to not talk about any cloud other than AWS; the second rule of re:Invent is… well you get the idea. But the pervasiveness of multi-cloud this year made it impossible to avoid. Businesses are investing heavily in other public/private clouds, stemming from a desire to minimize vendor lock-in and provide real choice to their users. Consequently, they want their cloud vendors to support more than AWS and provide migration recommendations at a multi-cloud and PaaS level.

Maturation of Management

Businesses need support for their maturing management model, which we at CloudHealth Technologies often refer to as “centralized governance and distributed management.” The CloudHealth platform helps by delivering best practice dashboards, reports, and policies to drive operational excellence, providing advice and guidance that helps customers be more efficient and secure.

Speaking of governance, many businesses are struggling to manage standardization, compliance, and automation in a scalable, centralized and auditable way. CHT’s own Chief Architect Vikram Pillai led a session on governance and policies during the show: Leveraging a Cloud Policy Framework – From Zero to Well Governed. Feel free to reach out to me for more information on this topic at @JoeKinsella.


Microservices was a hot topic of this conference. The industry is finally moving to a new application architecture; for almost 30 years we have been drafting off the client-server model. The next 30 years will be driven by microservice architectures, and clearly Amazon wants you to build all of this with AWS (“come on in, the water is great”).

So there you have it. Until next year!

Oh and if you didn’t get to swing by the CloudHealth Tech booth to try out our trivia game, here’s an online version you can use to test your KNOW IT ALL skills.