Day 3 Session Summaries
At the keynote on Day 3, there were so many machine learning product announcements that I am convinced that in a post-apocalyptic terminator-esque future if there are any historians left, they will look back on this day and realise that this was the inflection point where skynet went from being a mere possibility to an inevitability. Of course everyone was far too busy training their DeepRacers to notice or do anything about it. I keep looking over my shoulder expecting to see Arnold Shwarzenegger trying to get to the AWS Machine Learning team to destroy skynet before it happens.
Seriously though, lots of interesting Machine Learning features announced.
- dedicated ML chips (AWS Inferentia),
- Sage Maker; Ground Truth, a labelling service that combines humans and computers; and
- Sagemaker RL, Reinforcement Learning (which is what drives the DeepRacers everyone is playing with)
- Textract which is basically really intelligent OCR that can handle tabulated data and forms with smart context like birth date and SSN recognition
- AWS Personalize based on experience from Amazon personalization
- AWS Forecast
- AWS Optimized Tensor Flow
- Amazon elastic inference
Other announcements at the keynote were
- Glacier Deep Archive - for those times when Glacier cold storage is just not cold enough, and you want to try and save even more money.
- Amazon FSx Windows - Windows file system storage compatible with AD, and all the other windows file system goodness
- Amazon FSx for Luster - Dedicated for High Performance Computing workloads
- AWS Control tower - best practice blueprints and guard rails for setting up an AWS account
- AWS Security Hub - centrally managed security
- AWS Lake Formation - drag and drop (well almost) creation of data lakes
- Aurora Global Database - multi-region Aurora
- Dynamo DB Read/Write capacity on demand - auto scaling DynamoDB throughput. This is a real win for serverless application developers.
- Amazon Timestream - Intelligent storage of TimeStream data
- Quantum Ledger DB - cryptographically verified immutable ledger
- Amazon managed blockchain
- VMWare cloud - for when you just can’t get your workloads off VMWare
- AWS Outpost - for when you really need something on-premise, AWS will come and physically install racks of their servers in your data centre. They even come with the option of VMWare cloud
Like any good keynote it was part revival rally part marketing exercise, and you really need to be not to completely buy all of the shiney marketing spin. In some cases, this is AWS filling in some of the gaps they have between Google (e.g. dedicated ML chips) and Azure (e.g. Azure Stack), but it does stand to reason that AWS are still considered the thought leaders in cloud computing.
How to Build a Serverless Consulting Practice
This was a chalk talk focussed directly at consulting practices, so thought it important to go. The conversation basically centred around how to convince companies to consider companies to move workloads to serverless. The basic argument was that the reasons you should go serverless are the same as the reasons you should go cloud in the first place. Cost, convenience, reliability, scalability, availability etc… with the added mantra of “only worry about your own code, don’t worry about infrastructure”. The speakers also believed that as you start transition a business to serverless, you’ll begin to find yourself working your way up the business layers. Starting with moving simple cron jobs and log shipping operations you’ll be interacting mainly with developers, but when the management see the benefits of this approach, this should open up conversations about moving ETL or IoT ingestion workloads to a serverless architecture. Finally when the executive level see this, you’ll start to be able to work with them to solve real business critical problems, and help them towards a microservices architecture based on serverless.
At this point the conversation turned to the challenge of convincing organizations down the microservices path. This isn’t easy as microservices have had some bad press. Possibly because of poor implementations, possibly because of business structures being unwilling to change. Conclusions were
- Find the pain points, solve them in a determined manner with serverless
- First step is always the hardest, transformation will gather momentum after it is proven.
Another interesting thing to come up is that AWS are seeing traditional companies that are just coming to cloud embracing serverless more readily than companies already using cloud extensively. In essence leap frogging IAAS and PAAS architectures in favour of serverless.
Developing with .net core on aws: What’s new
This talk was a whirl-wind overview of a sample .Net serverless application that enabled the speakers to show off many of the .net integration tools and services provided by AWS. The application was a simple photo mosaic creator that allowed you to upload a photo library, then pick a single photo and the application would then create a mosaic that at a macro level looked like a grainy version of the photo selected, but when zoomed in was made up of individual photos from the photo library uploaded initially. This was achieved in 3 parts
- Library ingestion
- Front End
- Mosaic creator
Using services like Lambda, AWS Batch, EKS, S3, cognito. It demonstrated IDE tools in both Visual Studio and Visual Studio Code. The AWS Code* tools for build and deployment as well as an upcoming feature called the [AWS Mock Lambda test tool[(https://github.com/aws/aws-lambda-dotnet/tree/master/Tools/LambdaTestTool) which allows you to execute a lambda locally either from the deadletter queue or from trace input in either visual studio or visual studio code, and will support a proper F5 debug experience.
It was a great demonstration with a ton of code, which I hope is available somewhere. My one criticism is that they went through it so quickly, it left my head spinning.