Time for Andy's keynote - this year there are over 65,000 in attendance, and so left a lot of us to watch the keynote from overflow rooms or the live stream.
The theme for this years re:Invent is transformation. The cloud has always been a natural home for startups and innovators, but over the past 6 years the rate of adoption of the cloud in the enterprise has drastically grown.
Andy's keynote covered off 6 key dimensions, each connected with their own interpretation of some well known songs.
- Van Halen - Don't Want to Wait for Tomorrow
- Queen - Don't Stop me now
- Billy Joel - Movin out
- Dave Mathews Band - Too much
- Proclaimers – I'm gonna be (500 miles)
- Doors - Break on through
These 6 songs may have been stretching the connection to the point, but they did have some solid message and announcements underneath them.
Not just a technical change
The first (and by my view most important) topic of discussion was organisational support for change. Things are moving fast, and will only continue to increase. Throwing shiny technology at a problem without a plan is not only inefficient but also downright dangerous.
The 4 key factors to success are:
- Senior leadership team conviction and alignment
- Top-down aggressive goals
- Train your builders
- Don't let paralysis stop you before you start
In summary it reads like:
- Make sure there is support across the organsiation to push through the pain to transform out the other side.
- Do real things - pick some workloads and move them, find some problems and fix them - apply knowledge and skills together to achieve outcomes.
- Invest in your team - make sure they are trained and supported to take advantage of the changes that are available. Ensure you have modern support partners who can help you.
- Today is the day to act, don't let the size of the challenge stop you from starting. The team at Cevo can help you on your journey.
Once you get started, you don't want to be constrained by options. AWS tout the widest and deepest feature set on the market - and it starts with the core compute platform.
Lots of announcements in the Compute area:
- New ARM based Gravitron2 instance types (M6G, R6G, C6G)
- Inf1 Instances - ML inference engine instance types with custom ASIC
- Fargate for EKS
Andy flew over the Networking announcements, but from the screen grabs it shows a number of improvements to AWS Transit Gateway - the one that stands out is Multicast support. We will need to dig more into this over the rest of the week to see what it holds.
Moving into the cloud
The digs at Oracle continue this year and we also see a much more direct dig at Microsoft.
As part of the cloud migration section, Andy used a slide to depict moving to the cloud like moving home, and leaving some things behind in that journey. The cloud relocation is a great chance to review current systems and make some modernisation decisions to the technology landscape.
The 4 topics of discussion here were:
- Mainframe movement
- Looking beyond classic relational databases
- Closing windows
- Modern ISV & SaaS providers
There were no specific product launches during this section, I was hoping for some comments or additional tooling to the DMS suite to make things easier for customers making the switch.
A lot of discussion about appropriate tools for the right job, and not cramming all different data types into a single relational database scope - which leads right into the next section….
Data, Data, Data
When it comes to data - it is only increasing, so be careful what you wish for. AWS have been looking at the volume of data being produced and stored and have doubled down on helping solve these problems.
Data announcements are mostly around how to manage and control massive levels of data. Redshift is a key part to this - and its evolution to not just be the data control platform but access to additional data stores makes it a very compelling solution.
With Amazon Redshift you can now store data directly within it, join this data to S3 content from your DataLake using Lake House (Spectrum) and now join it with other 3rd party transactional systems using Federated Query.
On top of this AWS announced a new Query accelerator - AQUA - using the Nitro framework to move compute closer to the storage and building an intelligent caching layer.
ElasticSearch also got an uplift - a recognition that it is expensive to retain a full hot index of log data that is cost effective. The launch of UltraWark is a new storage layer for ElasticSearch that supports up to 3Pb worth of log data for a possible 90% cost savings. It uses a similar technique to AQUA to have more intelligence closer to the storage and automated cache management of what to have hot, and what to offload to S3.
- Release of S3 access points - make control and access to S3 simpler
- Redshift updates:
- UltraWarm for ElasticSearch
Needing to walk 500 miles
The proclaimers signaled the entry of the ML discussion, and a recognition from AWS that you still really do need to be a computer or data scientist to use their ML suite.
A number of developer experience improvements were announced, starting with a new IDE SageMaker Studio along with a number of efficiency improvements in the SageMaker tool chain.
- Amazon SageMaker Studio: The First Fully Integrated Development Environment For Machine Learning
- Amazon SageMaker Debugger – Debug Your Machine Learning Models
- Amazon SageMaker Model Monitor – Fully Managed Automatic Monitoring For Your Machine Learning Models
- Amazon SageMaker Processing – Fully Managed Data Processing and Model Evaluation
- Amazon SageMaker Autopilot – Automatically Create High-Quality Machine Learning Models With Full Control And Visibility
- Amazon SageMaker Experiments – Organize, Track And Compare Your Machine Learning Trainings
And we didn't stop there - while there has been a lot of investment in making the development of ML easier for the builders - there are a few new features here for those more from a buyer camp.
A new Amazon Fraud Detector service for financial services that builds upon Amazons extensive data set and allows additional customer supplied data to be modeled to provide a real time fraud detection solution.
Amazon CodeGuru looks to be the start of a new suite of tools to enter deeper into the development and operations workflow. Today marks the announcement of 2 different features - an ML powered Code Review tool. This tool integrates into your pull-request based development flow and provides feedback on sub-optimal coding practices or hard to spot problems. Will integrate with GitHub and CodeCommit on launch.
The second was CodeGuru profiler - this is an application runtime agent that profiles the application looking for expensive code paths and provides feedback to developers. It seems that amazon have finally put their toe in the APM market, something that will be interesting to see the impact on companies like NewRelic, AppDynamics and DataDog.
But still there was more - Contact Lens is a service pipeline that automatically extends the AWS Connect workflow to integrate sentiment analysis and engagement feedback reporting. I've had a number of discussions with customers over the past few months about adopting the AWS Connect suite, and one key item that holds them back is the technology enablement to get there. With Contact Lens, the value proposition for investing in the Connect platform is greater as you get the sentiment analysis outcomes without needing to do all the plumbing work.
And FINALLY in the ML section was the announcement of Amazon Kendra , an ML powered Enterprise Search solution that aims to solve the problem of locating key information items using the power of natural language search.
At Cevo we use a number of different technologies to store our internal assets, be that from Confluence or Google Drive - using Kendra aims to make the locating of the right asset much easier. When I get some spare time this week I hope to get Kendra connected to our data stores and see what insights it can bring for us.
Accessibility and Innovation
The last theme had a mixed focus, from VMWare to Verizon the announcements initially didn't seem to have a strong connection. But on reflection of what they were talking about these are all around innovating the way the cloud is consumed.
As the Cevo team discussed back in March (https://cevo.com.au/post/2019-03-02-the-future-of-the-cloud/) everything in computing runs in cycles, and what we centralise we are destined to decentralise again. The cloud has been a massive centralisation of compute, but with the 3 final announcements I think we are starting to see the turn into decentralisation.
AWS Outposts are now GA - so pull out your credit card and you too can own your own AWS compatible data center. They also announced a new VMWare Cloud option of outposts that allows you to retain your current VMWare toolchain, but gain access to the AWS Outposts services.
The announcements of AWS Local Zone recognises that compute must move closer to customers to overcome the speed of light issue. Amazon will be launching a number of smaller (and less resilient) data centers called Local Zones - these will have a limited set of AWS services and work much more like a single availability zone than a region.
And finally we get to AWS Wavelength - a partnership with Verizon to integrate AWS services directly into the edge of the telecom network to power the next generation of applications and IoT devices. No news on if there is an Australian partner for anything like this at this stage.
Lots of new and exciting announcements from Andy this morning. Recognition for all the hard working service teams at AWS - finally able to talk about their work and get it in the hands of a mass customer base. It will be interesting to see what Werner brings us later in the week.