AWS re: Invent’s main event each year is Keynote from AWS Global CEO Andy Jassy with no doubt. The speech will release the latest features of AWS services and the development trends of the future, so the live broadcast of the speech is also highly followed by global AWS enthusiasts and cloud users.
With the band playing, Andy Jassy began with sharing the development overview of the cloud in the past. He said that 97% of the workload situation is still on-premise, but in a 3% cloud workload situation, AWS occupied a leadership position of 47.8%. He then shared how AWS successfully transformed internally, including:
- Senior leadership team conviction and alignment
- Top-down aggressive goals
- Train your builder
- Don’t let paralysis stop you before you start
New Launches Announced at AWS re:Invent 2019
Andy Jassy continued releasing more than 20 new services, some of which are now available (GA), and some are still in preview. To make it easier for you to understand, we classify the new features into four major projects and explain how each project’s solutions to any problems that users encounter while using AWS, including Compute, Database / Storage, and Machine Learning and On-Premise Solution:
The new functions of computing service are mainly new architectures and specialized Instance types making the management and compatibility of container services easier and more diverse.
- New Instance types: M6g (Preview), R6g (Coming soon), C6g (Coming soon) backed by AWS graviton 2 (new ARM-based architecture CPU).
- New Instance type: lnf1 (GA), the first instance type designed by AWS specifically for Machine Learning.
- Fargate for Amazon EKS (GA): Support Kubernetes Service (EKS) on AWS to manage microservices on Fargate.
Continuing the theme of Database Freedom last year and deepening data warehousing related services. Strengthen the original service to the issue of authority control and cost optimization.
- Amazon S3 Access Point (GA): Enable you to define and control permissions when different teams and applications need to access S3 data.
- New features of Amazon Redshift: including Federated Query on data warehouse and data lake for query and analysis, and data lake export, which can direct the optimized data format of Amazon Redshift directly to S3.
- New Instance Type (GA): Fully managed storage optimized Redshift RA3 Instance can provide customers with faster and scalable data warehousing services.
- AQUA for Amazon Redshift (Coming in 2020): Compared with other cloud platforms, the new data warehouse hardware provides caching and can increase the reading speed by more than ten times.
- UltraWarm (GA): Amazon Elasticsearch service new warm storage tier can provide up to 900 TB of storage space, saving nearly 90% of the cost.
- Amazon Managed (Apache) Cassandra Service (MCS) (Preview): Fully managed Apache Cassandra service with high scalability and high availability.
AI / ML has been a hot topic in recent years. AWS has continuously introduced many new services in this area. The main new services focus on reducing machine learning barriers for users to enter ML.
1.AWS customized TensorFlow / MxNET / PyTorch Compared with the private machine running at the customer end, AWS customized ML Framework has improved performance by 20% / 22% / 22%.
2. The newly launched SageMaker Studio IDE solution lowers the barriers for users to use ML and provides more integrated features, including:
2-1 SageMaker Notebooks (Preview)：Traditional notebooks require a machine, and this new feature allows users to use the notebook without provisioning the machine.
2-2 SageMaker Experiments (GA)：Automatically save the entered parameters, settings, and results in SageMaker Studio, users can retrieve all the results in SageMaker Studio and compare them.
2-3 SageMaker Debugger：The devil is hidden in the details. ML’s biggest threshold is that if there is no clear insight during the entire ML build process, it will reduce the user experience of ML. AWS SageMaker Debugger Service (GA) is to provide a new integrated service to solve the problem that users can’t get effective insights from the stages of Build, Train, and Deploy in the ML process. Complete visual data and effectively provide model accuracy, reduce training time, and understand the process of model training.
2-4 SageMaker Model monitor (GA)：This service is used to detect the situation where the model has a concept drift in the production environment.
2-5 SageMaker Autopilot (GA)：As long as the service prepares the data, the Autopilot service can convert the data by itself, select the model that is most suitable for the algorithm and automatically calculate up to 50 different parameters for user review and comparison, and select the most suitable model to use. It provides AutoML that integrates full control and visualization data.
3. Amazon Fraud Detector (Preview)： Amazon Fraud detection has always been the most requested feature. Before this service is launched, users need to train this model by themselves. The biggest challenge for users is the accuracy of self-built models. AWS also launched a fraud detection API today, allowing users to directly use the ML API provided by AWS without having to create and train models themselves, to meet the needs of most users and reduce the effort required for self-construction.
4. Amazon CodeGuru(Preview)：The CodeGuru function utilizes multiple aspects of AWS internal development experience, AWS best practices, judging concurrency issues, judging incorrect processing procedures, and identifying correct input verification to assist users in code review.
5. Contact Lens for Amazon Connect (Preview)： For the Connect customer service system, Contact Lens can quickly analyze the semantic and effective text conversion of conversations in the voice in the customer service system and analyze the emotions of the calling customers, enhancing the user experience of the customer service system.
6. Amazon Kendra (Preview)： Use the machine learning function to read the information transferred by the user and provide the user behavior understood by machine learning. Users only need to directly enter the content of the website to create an enterprise-level search service.
For customers who have not yet fully moved to the cloud, the AWS has announced several key services specifically for this type of workload.
- AWS outpost is a fully-customized AWS hardware service announced last year at re:Invent 2018. It can send hardware directly to customers’ data centers, allowing customers to fully utilize native AWS services and consistency in their own data centers. Cloud user experience. This year, the service has been officially GA, and it can be used in the United States, Europe, the United Kingdom, Australia, Japan, South Korea, and other regions. Among the AWS services that can be executed are EC2, EBS, ECS, EKS, EMR, VPC, RDS (Preview), S3 (coming soon).Besides, OutPosts currently only supports AWS native services and has announced that it will support VMware in 2020.
- Amazon Local Zone (GA in LA)In response to whether customer applications need to provide low-latency performance for end-user units, AWS also launched Local Zone this year, which allows customers’ applications to be deployed in specific local zones, which are currently available in LA.
- Amazon Wavelength (coming in 2020)With the arrival of 5G, AWS today also invited the CEO of Verizon to talk about the future and evolution of 5G. To provide more innovative use cases, such as SDN, IoT and other mobile edge computing applications, AWS also announced the integration of Infra Entering the 5G network, the Wavelength service will be launched in 2020. It is expected that AWS Compute and Storage services will be integrated into the 5G network to provide single-digit low latency to Mobile Users and devices.
Nextlink on AWS re:Invent!
Andy Jassy also thanked AWS related partners, such as Nextlink, for helping customers build cloud environments that meet their needs. In the future, Nextlink will continue to maintain this spirit to create greater value for our customers.