2021 re:Invent, which releases should we pay attention to?

2021 Amazon Cloud Technology re:Invent has just come to an end, and many developers are discussing the heavy release of Amazon Cloud Technology re:Invent this year.

This is not surprising. Every year, Amazon Cloud Technology re:Invent releases dozens of new products and new features, some of which may represent the future development direction of cloud computing and the entire IT industry infrastructure, such as Amazon RedShift and Amazon Lambda, the former has guided the development of cloud-native data warehouses in the industry, and the latter has brought serverless into the attention of developers in the industry.

But this year's Amazon Cloud Technology re:Invent is somewhat different. In addition to the regular iteration of product performance, it also reflects the conceptual extension of cloud computing services. The most typical ones are Amazon IoT TwinMaker and Amazon IoT FleetWise.

Meta Universe and Internet of Things: The world is really clouding

According to the official introduction, Amazon IoT TwinMaker is a digital twin that allows developers to create real-world digital twins, such as buildings, factories, industrial equipment, and production lines, more easily and quickly. Through Amazon IoT TwinMaker, users can use digital twins to build applications that reflect the real world, improve operational efficiency and reduce downtime.

A digital twin is a virtual mapping of a physical system, which can be updated regularly according to the structure, state, and behavior of the real-world objects it represents. Amazon IoT TwinMaker allows developers to easily aggregate data from multiple sources, such as device sensors, cameras, and business applications, and combine these data to create a knowledge graph to model real-world environments.

Digital twin technology was first used for the health maintenance and protection of aerospace vehicles, and it was an unpopular technology. However, with the rise of the concept of "meta universe", digital twin technology has become more and more well-known, because the essential feature of digital twin is to map the physical world equivalently in the information world, so it has become one of the important supporting technologies for the concept of meta universe , Especially the industrial meta-universe, it has a wide range of applications.

The release of Amazon Cloud Technology re:Invent this time, Amazon IoT TwinMaker, can be said to be related to both Metaverse and Industrial Internet. In the past, the scope of cloud computing services focused on the Internet industry, facing the so-called traditional industries, mainly providing transformation services. But now, the service extension of cloud computing is rapidly expanding. Through the bridge of meta-universe, the scope has been expanded to the virtualized mapping of the entire physical world.

Closely related to this is Amazon IoT FleetWise. Amazon IoT FleetWise enables automakers to easily collect and manage data in any format (regardless of brand, model, or configuration) in the car, and standardize the data format to facilitate easy data analysis on the cloud. When the data enters the cloud, automakers can apply the data to remote diagnostic procedures for vehicles, analyze the health of the fleet, help automakers prevent potential recalls or safety issues, or improve autonomous driving through data analysis and machine learning And advanced assisted driving technologies.

If Amazon IoT TwinMaker provides a mapping service from the real world to the virtual world, then Amazon IoT FleetWise focuses on the field of Internet of Vehicles and solves the long-term development problems of Internet of Vehicles. The concept of the Internet of Vehicles first appeared in the 1960s, but for 60 years, it has been a bit "lame development". Most people understand the Internet of Vehicles to provide network services in the car instead of uploading vehicle data for analysis .

It was not until the birth of the Tesla Model S in 2012 that the Internet of Vehicles was included as a mandatory feature in the production line of automobiles. Now Amazon IoT FleetWise is released, bringing Internet of Vehicles related services to the cloud in an all-round way.

An important feature of Amazon IoT FleetWise is that it can build virtual representations of vehicles in the cloud and apply common data formats to build and label vehicle attributes, sensors, and signals. Amazon IoT FleetWise uses Vehicle Signal Specification (VSS) to standardize vehicle modeling so that signals such as "fuel pressure" are always expressed as fuel pressure and are measured in pounds-force per square inch (PSI) and kilopascals (kPa) . After the vehicle is modeled, upload a standard CAN database (DBC) or AUTOSAR XML (ARXML) file so that Amazon IoT FleetWise can read the unique proprietary data signal sent through the vehicle controller area network bus (CAN bus).

Understand, in fact, the underlying concepts of Amazon IoT TwinMaker and Amazon IoT FleetWise are exactly the same. They both build virtual maps in the cloud, but one is for the industrial field and the other is for the automotive industry. It can be said that the world is being virtualized, and it is also being clouded.

Amazon SageMaker Canvas: Create ML models without code

If Amazon IoT TwinMaker and Amazon IoT FleetWise embody the horizontal conceptual extension of cloud services, then Amazon SageMaker Canvas is the vertical conceptual extension.

Everyone knows Amazon SageMaker, as a fully managed machine learning service that has been released for four years. Amazon SageMaker provides developers with a complete "central kitchen". Developers using Amazon SageMaker only need to prepare "foodstuffs" (data) and start cooking (training models) directly, greatly improving developers and data scientists The efficiency of building, training, and deploying machine learning models has opened a new era of intelligence.

However, the AI field has long been restricted by the shortage of talents, and the application fields of AI are increasing, and the threshold for machine learning services needs to be further lowered. This is the purpose of Amazon Cloud Technology's release of Amazon SageMaker Canvas-to build machine learning models with code-free concepts and make model predictions to ensure that services can still be provided without the data engineering team. It uses the same technology as Amazon SageMaker to automatically clean and combine your data, create hundreds of models behind the scenes, select the best performing model, and generate new single or batch predictions. Supports multiple question types such as binary classification, multi-class classification, numerical regression, and time series forecasting.

There were many controversies about low-code and no-code in the industry before, but now it seems that this is not a concept dispute, but a real need in the industry. The release of Amazon SageMaker Canvas verifies this situation.

At the macro level of AI, whether it is based on AI-provided forecasting services or analysis services, it also breaks away from the pure pursuit of higher-level artificial intelligence, while taking into account the empowerment of AI capabilities to the industry. This is the further expansion and implementation of cloud computing's service concept.

Amazon Private 5G: Connect IoT devices with proprietary 5G

In this case, the release of Amazon Private 5G has naturally attracted many people's attention, because it is an important and necessary attempt to support service expansion. It can be said that Amazon Private 5G is one of the most important releases of this re:Invent.

On the mobile side, we have already used 5G communication services, but what companies need is a proprietary 5G service network. Amazon Private 5G can automatically set up and deploy the network, and expand the capacity as needed to support more devices and network traffic. David Brown, vice president of EC2 at Amazon Cloud Technology, said: “With Amazon Cloud Technology’s private 5G, we extend the hybrid infrastructure to our customers’ 5G networks to simplify, quickly and cheaply establish private 5G networks. Customers can start small, on-demand Expand, pay on demand, and monitor and manage their network from the Amazon Cloud Technology console."

Amazon Private 5G also focuses on serving the huge sensor and end-side device clusters based on Industry 4.0. The industrial meta-universe and the Internet of Vehicles mentioned above are naturally in the same sequence.

The world’s largest non-listed company-Koch Industrial Group (Koch) has reached a cooperation with Amazon Cloud Technology on Amazon Private 5G. The core of Koch Enterprise Group is petroleum and chemical, which is also representative of Amazon Cloud Technology. Service case.

Amazon Graviton3: The underlying computing power has been upgraded again

Of course, whether it is an IoT service such as Amazon IoT TwinMaker or Amazon Private 5G, it depends on the performance of the chip in the underlying instance. At this year's Yunqi Conference, Pingtou, a semiconductor company under Alibaba, released its self-developed cloud chip Etian 710 and announced that its performance exceeded Amazon Graviton2 by 20%.

From Amazon Graviton in 2019, Amazon Graviton2 in 2020 to Amazon Graviton3 today, Amazon Cloud Technology continues to improve computing services starting from the chip. Compared with Amazon Graviton2, Amazon Graviton3 integrates 55 billion transistors, single-core performance increased by more than 25%, floating point and encryption performance will increase twice . In terms of machine learning, Amazon Graviton3 includes support for bfloat 16 data, which will be able to provide up to 3 times the performance. the performance of 161b17141ea64c is leaping, the energy consumption of Amazon Graviton3 has been reduced by 60% compared to the previous generation.

The new generation of Amazon EC2 C7g instances are supported by Amazon Graviton3 processors and are also the world's first cloud computing instance that supports DDR5 content. Compared with the current generation of Amazon C6g instances supported by Amazon Graviton2 processors, the performance is improved by 25%.

Of course, Amazon Graviton3 is a general-purpose chip, and the dedicated chip has also been updated. Amazon Cloud Technology announced that a new Amazon Trn1 instance supported by Amazon Trainium, Amazon's second machine learning chip, will provide training for deep learning models in the cloud for use cases such as natural language processing (NLP), computer vision, search, recommendation, and ranking The best cost performance, compared with P4d instances, the cost of training deep learning models through Amazon Trn1 instances is reduced by up to 40%.

Amazon Nitro System chips also released new products. Amazon Nitro can be said to be a super black technology. To be precise, it is a set of architecture that can maximize the resources provided by the server to users and reduce virtualization losses. The so-called "virtualization loss" refers to the inevitable overhead in the network, storage, management and other system functions in order to maintain the normal operation of the service in the past. This overhead accounts for 30% of the overall server performance. The Nitro architecture focuses on these 30% of performance issues through customized hardware.

The Amazon Nitro System chip released this time mainly supports the Amazon EC2 Instance underlying management platform, which can share the workload for the CPU. General-purpose chips, special inference chips, and Amazon EC2 support chips are all neatly released this time. The Im4gn/Is4gen/ I4i instances of Amazon Nitro SSDs provide 30 TB of NVMe storage. Compared with the previous generation I3 instances, I/O latency is reduced by 60% and latency variability is reduced by 75%.

Data Serverless: The rapid promotion of serverless applications

Of course, in addition to the expansion of service capabilities at the network and end levels, the update at the serverless level is also worth noting.

The industry knows that Amazon Lambda has opened the era of Serverless, but it is still in 2019 that it has gained widespread approval and follow-up in the industry. This time, Amazon Cloud Technology re:Invent, Amazon Cloud Technology released four core product Serverless versions in one go: Amazon Redshift Serverless, Amazon EMR Serverless, Amazon MSK Serverless and Amazon Kinesis data streams on-demand.

Amazon Redshift, as we have mentioned, is the earliest cloud-native data warehouse; Amazon EMR is a Hadoop hosting service provided by Amazon Cloud Technology; Amazon MSK is a Kafka hosting service; and Amazon Kinesis data streams on-demand is a streaming data processing platform .

The serverless version of these services allows users to run applications built with these frameworks in a few clicks, without the need to configure, optimize, or protect the cluster.

Cloud big data architecture, because of the update of these serverless versions of Amazon Cloud Technology, the threshold is rapidly lowering. In the past, the construction of architectures like smart lake warehouses gave architects and engineers a headache, but now, the work of engineers is becoming a mere adjustment of parameters-Serverless's change to the industrial ecology is almost permanent.

Write at the end

Judging from the release of 2021 Amazon Cloud Technology re:Invent and this year’s cloud computing conferences, the product changes in the cloud computing field are one that focuses on the upgrade of the underlying basic computing power, which is a competition of hard core strength; the other is to focus on out-of-service. The key is how to understand the three roles of cloud, network and terminal and provide as common public cloud services as possible. The meta universe is a new concept emerging at the moment, providing new ideas and directions for the overall technological development direction, which deserves our special consideration.