Final Recap of Re:Invent 2017
by John Evdemon Tuesday, December 5, 2017

Key Observations:

1. Every popular service on AWS witnessed a new or a refurbished capability which clearly defines a path for next set of innovations to take place on the solution front for their customers.

2. Serverless, Data & Machine Learning stole the show from new service announcements standpoint, while Data, IoT, P3 Instances (GPU) and Deep Learning Frameworks were discussed in the context of re-imagined architecture patterns.

3. At this stage, AWS as a platform appears to have addressed all gaps where we appeared stronger than AWS, especially in the area of Machine Learning, IoT and Data Stack. Their popular services are getting stronger and stronger, fueling platform stickiness and it only gets harder for anyone to switch platforms going forward.

4. They are no more a platform who are perceived to provide stronger infrastructure services. On a comparative note, their managed services train appears to have just whisked past us considering the vastness and depth of announcements made at this conference.

5. Their investments in providing Machine Learning services catering to the likes of advanced ML practitioners, everyday developers and scientists is worth a note. Data & Machine Learning is undoubtedly the next big wave and their announcements just complimented this very stream.

Overall, their platform took a big leap on the ‘Managed Services’, welcomed ML Practitioners and Data Scientists with a very comprehensive ML offerings and went global scale with their Data & Analytics stack.

For further details check on the highlights below.

Highlights of the Event

AWS re:Invent 2017 announcements were focused on 3 key pillars and it was no surprise knowing the strategic direction they took starting last re:Invent. In fact, below are so significant that you can call them disruptive to an extent

1. Serverless

2. Data & Analytics

3. Machine Learning

In my view, the way the stories were told, this probably marks the beginning of a new wave that is going to redefine how software development happens on AWS. While Day 1 witnessed key announcements (Andy Jassy), Day 2 was predominantly on Architecture Patterns (Werner Vogels). Each keynote was 3 hours long.

Few stats presented during keynote

- 44K in-person and 60K live stream attendees

- AWS is now a $18 Billion run rate business with 42% Y/Y growth

- Largest # of Enterprise Customers, Public Sector, System Integrators, Academic and Non-Profit spread

- Special emphasis on ISV and SaaS provider ecosystem

- Tons of Customer References (Beamed 1000s of logos at every significant point as an evidence of vastness of their platform adoption)

- 60 New Services announced with Compute (Containers), Database & Machine Learning categories topping the list

Below summarizes key announcements under the 3 pillars

1. Serverless

No doubt ‘Lambda’ as a function/Event driven architecture has seen a hockey stick growth. But ‘Serverless’ as a pattern is beyond just Lambda. This was established during last re:invent and they re-established this year too. For all those who think Serverless means only Lambda, check out Serverless Computing page on AWS.

Key announcements in this category:

- AWS Aurora Serverless (Fully Managed – No need to provision instances, which is the case with Aurora today under RDS)

- AWS Serverless Repository

2. Data & Analytics

It was all about scale and flexibility. Global scale databases with multi-master and multi-region. They drove home the message using S3, Aurora, DynamoDB and Neptune (new Graph DB). Following were key to all the announcements in this space

- Making petabyte-scale analytics accessible to companies of all sizes

- Established S3 as their Data Lake which now supports SQL SELECTS

Key announcements in this category:

- Amazon Neptune

- Aurora Serverless and Multi-Master

- DynamoDB Global Tables (Fully Managed, Multi-Master, Multi-Region database)


3. Machine Learning

Significant time was spent in elaborating on the services that they are announcing in view of catering to the need of the industry. They layered their announcements under 3 buckets

a. Bottom Layer (Targeting advanced ML practitioners)

Their philosophy here is to support all Frameworks and interfaces.

- Frameworks -> Caffe2, CNTK, mxnet, PYTORCH, TensorFlow, torch

- Interfaces -> Keras, Gluon

b. Middle-Layer (Targeting Everyday Developers and Scientists)

- Announced AWS SageMaker

o Ability to easily build, train, and deploy ML models

o Baked in Top 10 most popular algorithms with flexibility to choose a framework

- Announced AWS DeepLens

o World’s first wireless, deep learning enabled video camera for developers

o Custom-designed deep learning interface engine. Interfaces with SageMaker and Lambda

c. Top Layer (For anyone who wish to leverage/integrate via APIs)

- ML Application Services (Lex, Poly, Rekognition, Comprehend, Transcribe, Translate)

Key announcements in this category:

- SageMaker

- DeepLens

- Rekognition Video

- Kinesis Video Streams

- Transcribe (Speech Recognition) and Translate (Language Translation)

- Comprehend (Managed NLP)

There were significant other announcements in the area of IoT and Compute. Launched their own Cloud IDE called Cloud9. Check out this page for all other announcements.