Although AWS re:Invent is virtual this year, there’s no shortage of content, sessions, and announcements. Last week, we covered major announcements from week one, along with CEO Andy Jassy’s keynote address. You can catch up on everything you might have missed from week one of AWS re:Invent 2020 in our articles here:
- AWS re:Invent 2020: Week One Highlights and Key Announcements
- AWS re:Invent 2020: Highlights from Andy Jassy’s AWS re:Invent Keynote
This week, we’ll cover week two of re:Invent, which focuses on what you need to know from the machine learning and infrastructure keynotes, as well as all the latest product announcements. To quickly jump to different sections of this article, click on these links: Machine Learning Keynote, Infrastructure Keynote, Key Announcements
Also, make sure you don’t miss out on our digital swag bag giveaway with $150 in AWS Cloud Credits, a complimentary Health Check on your environment, and a FREE 30-day trial of the CloudHealth platform!
AWS re:Invent 2020: Machine Learning Keynote
Swami Sivasubramanian, VP of Amazon Machine Learning, hosted the first-ever Machine Learning Keynote at AWS re:Invent, which included the latest developments in AWS machine learning, demos of new technology, and insights from customers.
Machine learning is no longer a niche solution. Swami explained that more than 100,000 customers use AWS machine learning to accelerate and improve their products and services, including companies like Domino’s, Roche, BMW, and Nike. In the keynote, Swami highlighted AWS’ commitment to providing machine learning services to enable the “freedom to invent” for customers and developers of all skill levels.
With this commitment in mind, AWS has released more than 250 new machine learning and artificial intelligence features in the last year (we’ll get to AWS’ most recent product announcements later on in this article).
Similar to Andy Jassy’s keynote, Swami also provided a list of key principles or “tenets” for success, but in the context of machine learning. Here are his five tenets:
1. Provide firm foundations
It’s important to start with strong frameworks and healthy infrastructure in order to set your teams up for success in training, development, and production.
2. Create the shortest path to success
Historically, machine learning development was a complex and costly process—and it still can be today. If you’re still relying on manual tasks and not leveraging the right tools to connect the dots between each step of the machine learning process, you’ll continue to be held back.
Swami shared several AWS services, including AWS SageMaker and AWS Data Wrangler, designed to help customers accelerate the adoption of machine learning so they can focus on solving real business problems, rather than managing data and infrastructure.
For all the sports fans out there (and even if you’re not), this portion of the keynote featured an interesting story in which Jennifer Langton, SVP of Player Health and Innovation for the NFL, explained how machine learning services are helping to predict and prevent player injuries by analyzing helmet safety data.
3. Expand machine learning to more builders
For Swami’s third tenet, he emphasized the value of expanding machine learning capabilities and services to more builders within your organization. Traditionally, non-experts have faced significant barriers to adopting machine learning, but it’s becoming more and more accessible—as it should! Good ideas and problem-solving can come from anywhere within the organization, so it’s important to equip all your builders with the power of machine learning to innovate and create change. Again, Swami shares specific AWS services to help support this tenet, including Amazon SageMaker Autopilot, Amazon Neptune, and Amazon Aurora for machine learning.
4. Solve real business problems, end to end
With any “wow” technology, such as machine learning, it can be easy to lose focus on all the things it can do and what you can build, and forget about solving real business and customer problems. Swami brings us back to the heart of technology, which is solving real problems end-to-end. He shares AWS customer examples (including Moderna’s “mission to develop a vaccine for COVID-19 and other life-threatening diseases”) and other practical AWS machine learning services.
For example, AWS Monitron is a new service that uses machine learning to detect abnormal machine behavior and enable predictive maintenance. Instead of manually monitoring equipment health, you can use AWS Monitron to automate this process.
5. Learn continuously
Swami’s final tenet is to learn continuously, which aligns well with the entire theme of AWS re:Invent—as individuals and businesses, we must continually observe, learn, and reinvent ourselves to succeed. AWS credits continuous listening and learning for how they’ve remained leaders in the cloud industry and at the forefront of disruptive technologies. So on that note, let’s keep learning and see what else re:Invent had to offer this week!
AWS re:Invent 2020: Infrastructure Keynote
Peter DeSantis, SVP of Global Infrastructure and Customer Support, took the stage this week for the Infrastructure Keynote with a focus on efforts to improve resiliency, availability, and sustainability for its customers.
A key theme for this presentation, and for AWS’ overall infrastructure strategy, draws from a quote from Werner Vogels, Amazon VP and CTO: “Everything fails.” Anything can and will fail, so it’s critical to anticipate failure and to build products and services designed to protect your customers.
AWS is taking steps in several areas, such as replacing central Uninterruptible Power Supply (UPSs) with “micro UPSs,” increasing the number of regions and Availability Zones, and optimizing the efficiency and performance of chips and processors.
New infrastructure design
DeSantis explained that if you want additional redundancy from a single UPS, adding another UPS to your infrastructure won’t solve all your problems, and could, in fact, create more risk.
To combat this, AWS is implementing a new infrastructure design that replaces single, large UPSs from their infrastructure with custom-built power supplies and small battery packs that are integrated into every data center rack. These micro UPSs are intended to reduce complexity and maintenance and improve availability.
Regions and Availability Zones
However, no data center, regardless of its architecture, is immune from failure. To improve availability and reduce latency for their customers, AWS continues to invest in more regions and Availability Zones (AZs) around the globe. Italy and South Africa launched earlier in 2020, and Indonesia, Japan, Spain, India, Switzerland, and Melbourne are in the works. The Los Angeles Local Zone is also available, with Boston, Houston, and Miami available for preview.
If you haven’t seen it already, I encourage you to explore AWS’ interactive global infrastructure site here, to see the various regions, AZs, local zones, and more around the world.
Amazon has been building custom hardware for a long time and has continued to evolve its offerings to increase performance, reduce customer costs, and increase security. DeSantis focused on three primary solutions during his keynote:
- AWS Nitro System: The AWS Nitro System is a combination of dedicated hardware and lightweight hypervisors that enable faster innovation and enhanced security. The most recent generation of instances built on the AWS Nitro System is the C6gn EC2 instances, which were announced last week during Andy Jassy’s keynote.
- AWS Inferentia: AWS Inferentia is Amazon’s first custom silicon chip designed to accelerate deep learning workloads and provide high-performance inference in the cloud. Inferentia offers up to 45% lower cost per inference compared to GPUs.
- AWS Graviton2: AWS Graviton2 processors power Amazon EC2 M6g, C6g, R6g, and most recently, T4g instances and their variants. DeSantis stated that Graviton2 is AWS’ most power-efficient processor, and is suitable for a variety of workloads, including application servers, micro-services, high-performance computing, electronic design automation, gaming, open-source databases, and in-memory caches.
Amazon has joined 30 other organizations worldwide in a commitment to achieve net zero carbon by 2040 as part of The Climate Pledge. DeSantis stated that AWS is focused on “improving efficiency in every aspect of our infrastructure.”
“From the highly available infrastructure that powers our servers, to the techniques we use to cool our data centers, to the innovative server designs we use to power our customer workloads, energy efficiency is a key part of our global infrastructure,” DeSantis added.
- Graviton2: Graviton2 are AWS’ most efficient processors, with up to 40% better price-performance than previous versions.
- New infrastructure design: AWS has seen 35% improved energy conversion by replacing the central UPS in their infrastructure with custom power supplies and small battery packs (“micro UPSs”) that are integrated into each data center rack.
- Renewable power: Amazon increased the amount of megawatts produced from renewable power by 300% in 2020, amounting to a total of 6,500 megawatts, sourced from 35 wind and solar farms around the globe.
Finally, I’d be remiss not to mention DeSantis’ competitor callouts regarding availability and resiliency. He presented several screenshots from Microsoft Azure and Google Cloud documentation, highlighting words that suggest unclarity or exclusivity, such as “usually,” “generally,” and “in select regions,” and compared that to how AWS describes the same. This was an interesting strategy—we’ll be interested to see if/how Microsoft or Google choose to respond.
AWS re:Invent 2020: Key announcements
Week one of AWS re:Invent 2020 already introduced a tremendous amount of new AWS products and services (see a recap of those here), but there’s still no shortage of announcements from week two. Here are the key announcements we’ve identified from week two of AWS re:Invent 2020:
- Amazon Redshift RA3.xlplus: RA3.xlplus is the third and smallest member of the RA3 node-family, creating even more compute sizing options for customers to choose from for their workload requirements.
- Amazon Redshift data sharing (preview): With data sharing, you can securely and easily share live data across Amazon Redshift clusters for read purposes without having to manually copy or move it.
- Amazon Redshift cluster relocation: Now you can move Amazon Redshift clusters between AWS Availability Zones (AZs) without requiring application changes.
- Automatic Table Optimization: Automatic Table Optimization is a new self-tuning capability that optimizes the physical design of tables by automatically setting sort and distribution keys to improve query speed.
- Amazon Redshift federated query (preview): Amazon Redshift’s federated query capability has now extended to Amazon RDS for MySQL and Amazon Aurora MySQL (preview), in addition to Amazon RDS PostgreSQL and Amazon Aurora PostgreSQL, which was announced as generally available earlier this year. Learn more about Amazon Redshift federated query here.
- Amazon Redshift ML (preview): Amazon Redshift ML makes it possible for data warehouse users to create, train, and deploy machine learning models using SQL commands.
- Native console integration (preview): Amazon Redshift now supports native integration with select AWS partners from within the Amazon Redshift Console. You can learn more about Amazon Redshift’s native console integration for partners here.
- Semistructured data support (preview): Available to preview, Amazon Redshift now supports native JSON and semistructured data processing.
- Amazon SageMaker Pipelines: New capability of Amazon SageMaker to build, manage, automate, and scale end-to-end machine learning workflows, enabling you to accelerate machine learning projects and scale up to thousands of models in production. Learn more about Amazon SageMaker Pipelines here.
- Amazon SageMaker Data Wrangler: With Amazon SageMaker Data Wrangler, you can simplify the process of data preparation and feature engineering, and complete each step of the data preparation workflow, including data selection, cleansing, exploration, and visualization from a single visual interface.
- Amazon SageMaker Clarify: Amazon SageMaker Clarify detects bias in machine learning models and increases transparency by helping to explain model behavior to stakeholders and customers.
- Amazon SageMaker JumpStart: Integrated with Amazon SageMaker Studio, Amazon SageMaker JumpStart accelerates machine learning workflows with one-click access to popular model collections and to end-to-end solutions that solve common use cases.
- Amazon SageMaker Feature Store: Amazon SageMaker Feature Store is a fully managed centralized repository for customers’ machine learning features, making it easy to securely store and retrieve features without having to manage infrastructure.
- Amazon SageMaker Debugger: Launched last year at AWS re:Invent, Amazon SageMaker Debugger automatically identifies complex issues that can arise in machine learning training jobs. Now, SageMaker Debugger can also profile machine learning models to enable customers to identify and fix training issues caused by hardware resource usage.
- Amazon SageMaker Data Parallelism (SDP): Amazon SageMaker now supports a new data parallelism library that helps machine learning teams train models on datasets with thousands of gigabytes, thereby increasing training times by up to 40%. You can learn more about SageMaker Data Parallelism here.
AWS Audit Manager
This is a new service that helps you continuously audit your AWS usage to simplify how you assess risk and compliance with regulations and industry standards. The service is available globally and offers a Free Tier so you can get started right away in the AWS Management Console. You can learn more about AWS Audit Manager here.
- Amazon Lookout for Metrics (preview): This new service uses machine learning to detect anomalies in customer metrics in order to proactively monitor business health, diagnose issues, and find new business opportunities. This is all with no machine learning experience needed for the end-user. You can learn more about Amazon Lookout for Metrics here.
- Amazon Lookout for Equipment (preview): Amazon Lookout for Equipment is an API-based machine learning service that detects abnormal behavior in equipment by analyzing data from customer sensors, such as pressure, flow rate, RPMs, temperature, power, and more.
- Amazon Lookout for Vision (preview): Amazon Lookout for Vision is a new fully managed machine learning service that detects visual defects on production units and equipment.
We hope you found this helpful! Enjoy the final week of AWS re:Invent 2020 starting on December 14th, and if you haven’t already, make sure you stop by the CloudHealth virtual booth for a product demo that shows how we help you get the most of your AWS investments, including support for AWS Savings Plans and CloudHealth Secure State—a solution for mitigating security and compliance risk and much more.
Also, keep an eye on the CloudHealth blog, where we’ll be providing a final recap and our key takeaways from the final week of AWS re:Invent 2020.