Re:Invent 2019 Part 2
As I mentioned in Part 1, attendance was high. Now in its 8th year, Re:Invent had the capacity for 65,000 with over 3,000 sessions.
This year, “Transformation” was the main theme. The main take away was that you should be moving more aggressively to the cloud. Andy Jassy in his keynote kept coming back to this theme, and much of it was aimed at senior leadership. He laid out some key points to accelerating the journey.
Senior leadership Team conviction and Alignment
Top-down aggressive goals
Train your builders
Don’t let paralysis stop you before you start
While it all sounded good, I think Jassy has a bit of a pollyanna worldview when it comes to the cloud. He did make a great point that cloud innovates infrastructure faster than on-prem can, and I think this is ultimately the reason cloud will win out in the end.
One of the big points AWS was making is that not only do they have the broadest range of products, they also have the greatest depth in those categories. Jassy noted that they have over 175 products, while I count as many as 191 listed on their website. Although I would take this number with a bit of a grain of salt, since many of the “products” seem more like features and capabilities of a product than an individual product. He also managed to make his usual digs at Oracle, but this year, he added in MSSQL to the mix saying that people are racing to replace the Microsoft database. I think this speaks to the strength of Azure’s recent growth. Jassy said that 50% of Windows in cloud run on AWS and he referenced IDC estimates that by 2020 80% of workloads will be deployed on Linux.
As always there were a ton of announcements. On the compute side, in addition to offering up Intel, AMD and ARM instances, they have been busy building a lot of custom chips. With their acquisition of Annapurna Labs, they have grown their chip design capabilities. Their line of Nitro chips brings about the promise of disaggregating the CPU. Their plan is to offload key functionality to customers silicon to drastically accelerate functionality while creating an opportunity to scale massively, while isolating security problems.
AWS Announced M6G, R6G, C6G, and AWS Graviton 2 instances. 80-90% of the costs of ML are associated with inferencing, and to help with this they unveiled Inf1 instances. These Inf1 Inferentia instances offer low latency, 3x Throughput and are Integrated with all the major frameworks.
81% of all cloud containers run on AWS, and now they have extended the popular Fargate managed service to Kubernetes with Fargate EKS.
There are a bunch of new things on the network side with AWS transit gateway multicast, AWS Accelerated Site to Site VPN, AWS transit gateway inter-region peering, and AWS transit gateway network manager. On the security security side they gave us IAM Access Analyzer and granular access control for Amazon S3 access points.
AWS offers a lot of database options, and they are clearly proud of this. On the data warehouse front, they made over 100 updates to Redshift, and they now are offering federated queries across many of their products. In addition to RA3 instances which allow you to scale compute and storage separately, they are utilizing their Nitro chips to offload some of the compute to the raw storage which can improve performance and reduce costs with their Aqua for Redshift offering. They also introduced Datalake Export to simplify integrating data back into a data lake as well as Ultra Warm storage tiering solution
On the ML front there were a bunch of announcements as well. They announce the web-based IDE (integrated development environment) SageMaker Studio, SageMaker Experiments which simplifies saving all your experiments and artifacts, SageMaker Debugger which helps you interpret models, and SageMaker Model Monitor, which helps you monitor for concept drift.
They also are trying to tackle the Auto ML challenge with SageMaker Autopilot which takes CSV Training data, automatically transforms the data, selects best algorithm, trains up to 50 different models, allows you to Inspect and compare models and ultimately deploy the optimal model.
They brought some core functionality from their retail business to the AWS world with Amazon Fraud Detector an ML based fraud detection tool, and Contact Lens an ML powered contact center analytics. Amazon Kendra provides Machine Learning powered Natural Language Processing search for the enterprise.
Previously announce AWS Outposts have now hit general availability with
EC2
EBS
ECS
EKS
EMR
VPC
RDS (Coming Soon)
S3 (Coming Soon)
This comes in AWS Native or VMWare Cloud on AWS which arrives in early 2020. They are leveraging Outpost to create Local Zones Amazon managed buildings to move the cloud closer to the business. This will help them achieve single digit millisecond latency, and is available by invitation only at this point.
Lastly, they are jumping on the 5G bandwagon with AWS Wavelength. Through a partnership with Verizon, they are offering single digit millisecond latency edge applications with M6G, R6G, C6G, Graviton 2, and Inf1 Instances. This is currently in preview in Chicago.
That is a lot of stuff, and if you dig deeper there is more, but one can only fit so much in a video. So thanks for watching, I very much appreciate your likes, so if you feel so inclined pleas like, subscribe and hit that bell for regular updates. And, I will se you in the next video.
Announcement Overview
Instances
M6G, R6G, C6G AWS Graviton 2
Inf1 Instances
Nitro Enclaves
Containers
Amazon Fargate for EKS
Network
AWS transit gateway multicast
AWS Accelerated Site to Site VPN
AWS transit gateway inter-region peering
AWS transit gateway network manager
IAM access Analyzer
Amazon S3 Access Points (granular access control)
Federated Query
Redshift, S3 and relational database services
Datalake Export
Get data back into the data lake
Redshift RA3 instances
Allows you to scale Compute and Storage separately
Tiers data to S3 with High speed connectivity
Aqua for Redshift (advanced Query Accelerator)
Compute on raw storage without having to move it
Ultra Warm (warm Tier on Steroids)
Accelerated Warm Storage Tier
Storage Tiering Solution
Could save up to 90%
Amazon Managed Cassandra Service
SageMaker Studio
Web Based IDE (integrated development environment)
SageMaker Experiments
Save all your experiments and artifacts
SageMaker Debugger
Transparency for Models
Helps you interpret models
SageMaker Model Monitor
Monitors for Concept Drift
SageMaker Autopilot
Input CSV Training data
Automatically transforms data
Selects best algorithm
Trains up to 50 different models
View Model
Inspect and compare models
Deploy model
Auto ML
Amazon Fraud Detector
ML fraud detection
Contact Lens
ML powered contact center analytics
Sentiment analysis
Makes contacts searchable
Amazon Kendra
ML powered NLP search for the enterprise
AWS Outposts general availability
EC2
EBS
ECS
EKS
EMR
VPC
RDS (Coming Soon)
S3 (Coming Soon)
Fully managed solution
AWS Native or VMWare Cloud on AWS (early 2020)
Local Zones
General Availability by invitation
Single digit millisecond latency
Amazon managed buildings with Outposts
AWS Wavelength
Single Digit millisecond Latency Edge Applications
First partnership with Verizon
In Preview in Chicago