Andy Jassy, CEO of Amazon Web Services (AWS), took the stage Tuesday morning to kick off day 2 of re:Invent 2019, where he shared insights into the industry’s latest trends and challenges, showcased some of AWS’ most innovative customers, and announced numerous new AWS products and services aimed at helping customers transform their business.
Whether you’re one of the 65,000 lucky enough to be experiencing all of the action first-hand in Las Vegas or one of the many thousands of viewers streaming sessions from your less-stimulating—but arguably much more comfortable—desk or couch, Jassy had the same goal in mind for all of us: transformation.
How should organizations think about transforming themselves in order to stay meaningful and sustainable over time? How can organizations reinvent their business to stay innovative and continuously adapt to emerging challenges? AWS thinks they have the answer—and they’ve doubled-down on product innovation over the last few years to prove it.
Here are all of the most transformative announcements from Andy Jassy’s re:Invent 2019 keynote:
Innovation starts with instances
While AWS continues to push the boundaries on innovative services like machine learning and 5G (read on for exciting new announcements on those fronts), they haven’t forgotten about their more basic, and some would argue less sexy, technologies like ECS instances. In fact, AWS is offerings 4x more instance types than this time 2 years ago. But what exactly is it about these new instance types that make them so different than previous iterations?
Jassy points to a focus on reinventing technologies fundamental to running instances: the hypervisor (with AWS Nitro System) and chip technologies (through the acquisition of Annapurna Labs from Israel and partnership with Intel, AMD where teams were tasked with increasing chip capabilities while reducing costs).
New Instance Announcements:
- M6g, R6g, and C6g Instances for EC2. Powered by arm-based AWS Graviton2 processors, these new instances offer 4x more compute, 5x faster memory, and 7x times better performance than the first-generation of Graviton processors. What’s the bottom line for AWS customers? Almost 40% better price-performance than first-generation Graviton processors. (Learn more) Note that M6g is available today, with R6g and C6g availability coming early 2020.
- Inf1 Instances for EC2. Powered by AWS Inferential chips, AWS claims Inf1 instances offer the highest throughput (3x faster than the next leading), cost-per-inference instances in the cloud. (Learn more)
Let’s not forget about containers
AWS claims to offer the best (and broadest choice) for running containers, with Amazon Elastic Container Service (ECS), Amazon Elastic Kubernetes Service (EKS), and AWS Fargate. And with 85% of Kubernetes running on AWS, they’re probably right.
Fargate is the only compute engine that allows users to focus on managing containers without having to worry about provisioning and managing servers. Since it’s release, Fargate has seen exponential adoption due to its ease of use, with over 40% of new AWS container users starting their container journey with Fargate according to Jassy. Because of the overwhelmingly positive response to Fargate for Amazon ECS, AWS decided to expand Fargate’s capabilities…
New Container Announcements:
- Amazon EKS on AWS Fargate. Kubernetes customers now have access to all of the benefits of running containers in AWS based on serverless technology of Fargate. (Learn more)
AWS tackles big data head-on
One space in which Jassy says companies need to think about transforming themselves in is the world of big data. Regardless of your industry, gathering, making sense of, and effectively drawing insights from your data is extremely difficult (as well as time-consuming and expensive). Even with some of the tools and services AWS currently offers to help tackle big data, challenges still remain.
New Data & Storage Announcements:
- Amazon S3 Access Points. Simplify data access management at scale for applications using shared data sets on S3. Easily customize access permission rules for each of your applications, restrict access to within your private networks, and create hundreds of access points on a single bucket. (Learn more)
- Data Lake Export Query. Unload data from a Redshift cluster to S3 in an efficient open columnar storage format optimized for analytics and query across complete data lakes. (Learn more)
- Amazon Redshift RA3 Instances with Managed Storage. Scale storage and compute separately. When required, less frequently accessed data will move to S3 automatically. (Learn more)
- AQUA (Advanced Query Accelerator) for Amazon Redshift. AQUA brings compute to the storage layer with 10x better query performance than any other cloud data warehouse (and is 100% compatible with current versions of Redshift). (Learn more)
- UltraWarm. Storing data is expensive at scale, often limiting the amount of data retained in analysis and impacting the depth of available insights. UltraWarm, the new warm storage tier for Amazon Elasticsearch Service, reduces costs by 90% (with an equivalent volume of data), can scale up to 3 PB of log data per cluster, and can analyze years of operational data. (Learn more)
- Amazon Managed Apache Cassandra Service. Managing databases at scale is never easy. Amazon MCS is scalable, highly available, and managed Apache Cassandra-compatible database service. Amazon MCS offers single-digit millisecond performance, cluster-free management, and pay-as-you-go resource usage. (Learn more)
Huge strides in machine learning capabilities
AWS is arguably the best place to run machine learning, supporting all 3 of the major ML frameworks: TensorFlow (85% of TensorFlow running in the cloud runs on AWS), PyTorch, and mxnet. With the release of Amazon SageMaker in 2017, AWS removed the heavy lifting from each step of the machine learning process to make it easier to develop high-quality models. While there have been over 50+ capabilities added to SageMaker in the past year alone, it’s still challenging for customers to gain visibility into their ML models—sending AWS back to the drawing board to drum up more solutions.
New Machine Learning Announcements:
- Amazon SageMaker Studio. The first fully integrated, web-based IDE for complete machine learning workflows. Developers can now build, train, and deploy their models in a single interface. (Learn more)
- Amazon SageMaker Notebooks. Open notebooks in seconds with a single click, without the need to provision instances. Increase or decrease compute resources without interruption and automatically copy and transfer Notebook content to new instances. (Learn more)
- Amazon SageMaker Experiments. Saved by SageMaker Studio as an experiment, easily capture, organize, and search your experimental models in one place. (Learn more)
- Amazon SageMaker Debugger. Improve the accuracy of your ML models through automation. The debugger is on by default and provides real-time metrics of model training performances which can be viewed through notebooks in SageMaker Studio. (Learn more)
- Amazon SageMaker Model Monitor. Detect concept drift by automatically monitoring models deployed to production and alert developers when a drift is detected. (Learn more)
- Amazon SageMaker Autopilot. This is auto-machine learning for your ML models. Train up to 50 models at any given time, inspect and compare model performance, and choose the most appropriate model for production. (Learn more)
- Amazon CodeGuru. CodeGuru is aimed at helping developers write and commit code with minimal errors. CodeGuru pulls and reviews code to provide an assessment that, among other things, can detect anomalies, concurrency issues, and incorrect inputs. CodeGuru helps to enhance the code using machine learning by providing suggestions to replace the most expensive lines of codes to save money and improve the CPU utilization. (Learn more)
Reaching customers on-premises, in cities, and with 5G
At last year’s re:Invent, Jassy announced an industry-changing partnership with VMware, releasing VMware Cloud on AWS to help on-premises customers gain access to innovative cloud products and services while allowing them to continue using tools and services they were familiar with on-premises. Following the theme of on-prem, Jassy also announced the team was working on AWS Outposts, a service that would run AWS infrastructure on-prem for a truly consistent hybrid experience.
Today, Jassy announced the general availability of AWS Outposts, with future support for AWS RDS and S3. (Learn more)
Jassy also talked about the challenges consumers face with latency in large metropolitan areas. How can AWS offer end-to-end support for all consumers and applications, regardless of where they are being used?
Announced today, AWS Local Zones. A new type of AWS infrastructure deployment will deliver single-digit millisecond latency to end-users by placing compute, storage, and database services close to large metropolitan areas. At this time, general availability is only in Los Angeles, with more cities to follow. (Learn more)
After challenging the AWS team to develop innovative services that can reach customers on-premises and in large metro areas, Jassy highlighted an 18-month partnership with Verizon, where the two providers worked together to reimagine the virtualized network in order to push the boundaries of 5G.
Announced today, AWS Wavelength. This new virtualized network brings single-digit latency to mobile devices and users with AWS compute and storage at the edge of the 5G network. (Learn more)
New AWS announcement rivals Microsoft’s Project Cortex
To round out his re:Invent 2019 keynote, Jassy continued to focus on the ways in which machine learning can help drastically transform an organization’s operational efficiency and customer satisfaction—both of which are crucial to helping a business reach success.
One way machine learning can be of particular use is through transcription services (if your brain immediately went to call center logs, you’re right there with Jassy). For organizations around the world, across different industries, customer satisfaction is made or broken on customer service calls.
Today, Jassy announced Contact Lens for Amazon Connect (learn more). Machine learning-powered contact center analytics for Amazon Connect can help users transcribe and analyze vast amounts of customer calls (in addition to those calls that have previously been recorded). With Contact Lens for Amazon Connect, call logs can be analyzed to identify things like emotions and sentiments, recognize and flag long periods of silence (which can be indictative that a call center representative is ill-informed or unable to access correct information), and note when people are talking over each other. These insights can help businesses gain clarity around the state of the call centers (and by association customer satisfaction) and make informed decisions around how to improve the customer service process.
But if AWS can use machine learning to search and analyze call center transcription logs for specific words and phrases, can the same type of machine learning search services be applied to help large enterprises improve their internal knowledge base and document library? That’s what customers were asking, and AWS and Jassy delivered.
Announced today, Amazon Kendra. With Amazon Kendra, AWS customers can reinvent their internal enterprise search capabilities. With Amazon Kendra, easily pull data from different silos across your organization like Sharepoint, ServiceNow, HDD, etc., and sync and index this data using machine learning for later resurfacing during internal search. Amazon Kendra is a web-based console that integrates with your own website.
Amazon Kendra rivals Microsoft Azure’s recent announcement of Project Cortex (announced at Microsoft Ignite just one month ago), as both providers look to place their machine learning search functionality directly into the databases of large enterprise organizations.
With all of the announcements to unpack from Jassy’s keynote plus those to come throughout the week, you’ve got a lot to unpack. Join our product experts as they review all the key announcements from re:Invent 2019.