Home > AWS, re:Invent > AWS re:Invent 2017: The Day 3 Buzz

AWS re:Invent 2017: The Day 3 Buzz

November 30th, 2017

2017-11-29 07.09.04

Andy Jassy’s Keynote

2017-11-29 08.00.51An early wake up to get to the keynote in time with the hordes. Common sense would say to rather watch the stream from the comfort of somewhere else but I wanted to get a sense of the vibe and how the inevitable kool-aid gushing felt “live” as it was delivered.

I’m not going to live blog the keynote, way too much for that and it will be covered extensively elsewhere particularly by Scott Lowe: AWS re:Invent 2017 Keynote with Andy Jassy

Andy Jassy had a fun way to go through some of the themes by having a house band on stage playing 5 songs The words were on the BIG screen and he used song lyrics to explain how AWS “thinks”.

There were SO many announcements at some stages I actually felt a little overwhelmed, not from the number of announcements but more for the implications of what we’re seeing.

Wake Up Enterprise and smell the Cloud

We are seeing an incredible amount of service building by AWS itself to create some pretty impressive capabilities. I’d go far enough to say that if you’re thinking of installing or running any sort of platform yourself be it infrastructure, platform, database, analytics, containers…just don’t bother.

From the announcements today this is extending to machine learning, natural language processing and IoT. Setting up any of these environments yourself is a fools game. Get over yourself and your empire building. Enterprises are dying to take advantage of machine learning but they generally cannot store and process the data themselves as its just too hard. AWS just takes this all away and it gives you amazing power.

Cloud Power

This is the true power of public cloud, its all the clever services you can tap into that you just can’t do as well or at all yourself. Reasons for resisting are non existent and you’re doing your company and customers a disservice by again not getting over yourself.

I’m certainly not being an AWS shill here, Azure and GCP also have very many compelling offerings and I don’t intend to compare them. I’m at re:Invent though so my focus today is on AWS.

The inertia that is stopping you going all in on public cloud (whichever one or many) needs to stop. You need to start migrating your services to public cloud now even if its lift and shift and then start the refactoring and sorting out your current rubbish. You need to seriously invest in people training and get on board. Why you ask? “We don’t know how much it will cost and our internal processes need to be refactored.” Yes, you need to learn about cloud costing, it may actually cost you more initially but you will have orders of magnitude more super powers (I can’t think of a better way now to label this!). The opportunity cost however can be very serious.

Look at all  the current AWS services, look how many there are which you can tap into and more importantly don’t have to reinvent yourself.


What can you do TODAY

Just think of what you could turn on TODAY based on what was announced if you have all or even some of your data in AWS. Take the language stuff. You can Transcribe speech into grammatically correct text starting with English and Spanish. All your call center calls can be converted to text for other analysis in real time. Translate does real-time language translation as well as batch translation and soon with automatic language support. Think of how you can expand your customer base across the world and use language features to be able to service them.

Comprehend looks at text information and can do analysis and identifies things like people, places, phrases, sentiment and language and this can be used across large sets of documents. You can point Comprehend at all your call center text logs and pull out so much information which could be used to detect issues as they happen for future decision making. You can get the sense of every chat session in use or see if your employees are angry at something. You can pull in social media feeds and in real time work out if your keynote messaging is working and maybe even tweak things on the fly.

How long would it take to do this yourself. There’s no way, having your gravity of data in the cloud makes these kinds of things possible.

Advancing Infrastructure

Also think of the infrastructure platform enhancements you can take advantage of today or soon. Managed databases are so important to your business as its all about state and can’t be started/stopped on the fly. Its where you store your business information. How long does it take to set up a SQL server and do log shipping or availability groups? Yeh, not that quick and certainly not globally scalable. If you are public clouding and using native cloud databases like Aurora for MySQL/PostgreSQL, you can soon have multi-master Aurora with writes across AZs and soon regions. Same with DynamoDB with Global Tables, multi-write, multi-region key-value store. You can now extend your application globally with users from across the world accessing the same database with local latency. You had to do a large amount of manual hand cranking to do this before even if already using Aurora or DynampDB, now its just a platform feature update, that’s pretty impressive infrastructure.

The AWS Sensing Engine

Remember, the bigger the AWS platform, the more AWS can see what’s going on, the more it can foresee what customers need. They throw out the line “most customer obsessed company” which at first glance doesn’t sound genuine. They can do this by listening to customers which most other tech companies also do but they also have the data. They can see exactly how many customers are replicating DynamoDB tables from one region to another and can then decide to make it easier.

Amazon has this incredible sensing engine that mines its own data to build new services and its going it be able to see ALL customers and create services they need.


2017-11-29 08.35.38Also so much focus on this. AWS deliberately don’t equate serverless with Lambda as many others do but rather a suite of services where you don’t have to manage an OS with its patching and maintenance. This is very broad as it includes things like S3, CloudFront, API Gateway etc. Lambda is a subset of Functions as a Service (Faas) which is a subset of serverless. I’ve been speaking at VMUGs recently, talking about serverless and how infrastructure is evolving and probably need to update my talk and explain this a little better Functions is a part of the whole suite of things you manage that you don’t have to worry about an OS.

Going after the Enterprise

The reality is enterprises have a way to go to organisationally be able to take advantage of all the toys. I had been wondering how much AWS would need to bend towards the needs of enterprises. I’m now not sure whether AWS sees a need to do this.

They have a long term vision of how IT services should be delivered and they’ve been proven very right so far. They are going to create compelling services for all their customers and enterprises can chose to take advantage of them. Many services are practically impossible for enterprises to build or even buy themselves internally. AWS doesn’t need to target enterprises as hard as most other IT vendors as they know they’ll come running anyway. Ponder that for a moment. Training and skills is obviously a big barrier to this so I see AWS putting more resources into training to get the new cloud builders on board.

AWS Greengrass on vSphere Preview

An interesting new option from VMware announced today is AWS Greengrass on vSphere. If VMware on AWS is about running VMware workloads on/in AWS, then this announcement is just the reverse, running some AWS workloads on VMware with all the places that can run.

I had wondered whether something like this was a possibility and was sure there was some agreement that in order for VMware to want to partner with AWS there was going to be something more for VMware’s benefit for traditional on-prem workloads from AWS.

This is an IoT play where you may need local data processed with quicker transactions times than heading to the cloud.

An AWS Greengrass vSphere appliance is deployed from an OVA. It runs an Ubuntu VM which runs a “Greengrass Core”. This then connects back to AWS which you manage from the AWS console. Lambda functions can then run on this appliance.  I wonder how much this can be extended, is it only IoT, it uses a MQTT message bus, can you use this as a generic message bus and get Lambda to see this and fire off events as non IoT messages appear? This could be hugely beneficial to bring Lambda functionality on-prem.

Read more https://blogs.vmware.com/vsphere/2017/11/aws-greengrass-on-vsphere.html

WIN313 Ensuring Your Windows Workloads are Well Architected

Julien Lepine from AWS

A dash across Vegas and another missed walk-up place even 30 minutes before.

My reason for wanting to do this session was that AWS may be very Linux focused but I’ve been a Windows person for every so wanted to hear about specific Windows things for AWS.

Based on the synopsis, Julian was going to go through the Well-Architected Framework which is all about how to best design your workloads on AWS. AWS has recently introduced overlay documents for it, one of them being for Windows.

The reference architectures and best practices for Windows have been described in this paper.


ENT346  —  How AWS Runs Our Weekly Operations Meetings

David Lubell from AWS

I was really looking forward to this session, as this is the very first time AWS has opened the kimono on how it actually runs its own operations.

I wrote about it separately How AWS Runs Our Weekly Operations Meetings – ENT346


Time for the social activities to kick in which is a busy night

Serverless Happy Hour

2017-11-29 16.42.20 2017-11-29 16.41.53

Pub crawl and Tatonka Challenge

2017-11-29 17.48.18 2017-11-29 17.48.27

2017-11-29 18.57.07 2017-11-29 17.55.23

2017-11-29 17.55.29 2017-11-29 18.48.36

Cloud Beers

2017-11-29 19.33.00 2017-11-29 19.28.49

Rubrik re:Invent Party

Good to meet up with many people at the Rubrik Party, Justin Warren, Emad Younis, Rebecca Fitzhugh, Chris Wahl, Bryan Krausen and Eric Shanks.

 2017-11-29 22.14.19

Categories: AWS, re:Invent Tags: , ,
Comments are closed.