Home > News content

Amazon AWS released more than 10 new products in 180 minutes: machine learning chips, block chain services

via:博客园     time:2018/11/29 19:25:36     readed:496

As soon as AWS's CEO Andy Jassy stepped on the stage, there was a long whistling and applause


AWS re: Invent, which was held in Las Vegas at 8 a.m. on November 28, officially entered the agenda of the third day. Today is a particularly busy day, with a long queue at the entrance to the venue from 6:30, which depends largely on Andy Jassy, CEO of Amazon AWS today, bringing keynotes lasting three hours.

Sure enough, Andy Jassy has released more than a dozen blockbuster products and services in 180 minutes (how does this compare with Apple's launch?) It mainly includes machine learning chip, block chain, storage, database, machine learning and hybrid cloud.


At the outset, of course, it's about performance and market share. Q3 revenue in 2018 was $27 billion, an increase of 46%.


Global cloud computing market share chart, by the way to Oracle (live burst of laughter. Prior to Oracle CEO Larry

Amazon uses the re: Invent conference every year to highlight new tools and functions, and this year seems no exception. Lei Feng's site for you to interpret.

A List of 15 Heavy New Products

  • Ultra-low-cost cold storage on the Amazon S3 Glacier Deep Archive cloud.

  • Amazon FSx for Windows File Server is fully hosted on a Windows Windows Windows Windows Windows Windows Windows Windows Windows File Server.

  • Amazon was FSx for Lustre Fully Hosted Lister File System for Heavy Computing Load.

  • Amazon Control Tower provides a secure and compliant multi-account environment as the Landing Zone landing area.

  • AWS Security Hub centrally manages security and compliance services in an AWS environment.

  • Amazon Lake Formation helps you build a secure data Lake in a few days.

  • Amazon Timestream is a fully managed, fast, scalable time series database.

  • Amazon Quantum Ledger Database (QLDB) is a fully hosted distributed accounting database owned by the central authority, providing transparent, unchangeable and encrypted transaction authentication.

  • Amazon Managed Blockchain fully hosted block chain services supporting Super Accounting Platform Hyperledger Fabric and Ethereum architecture.

  • Amazon Elastic Inference assists graphics acceleration processing on EC2 instances to achieve low-cost and fast inference operations.

  • Amazon Marketplace for Machine Learning provides hundreds of machine learning algorithms and models that can be deployed directly on Amazon SageMaker.

  • Amazon Textract is an optical character recognition service that extracts text and data from almost any file. It can be used without machine learning experience.

  • Amazon Personalize provides real-time personalized customization and recommendation services based on the same technology used by Amazon.com. It can be used without machine learning experience.

  • Amazon Forecast provides accurate time series prediction services based on the same technology used by Amazon.com. It can be used without machine learning experience.

  • AWS Outposts runs the AWS infrastructure in the local data center and achieves a consistent hybrid architecture experience.

Next, weSeveral important releases will be detailed for you

AWS launches a new Inferentia machine learning chip

Encouragingly, Andy Jassy announced a new dedicated machine learning chip called Inferentia on the spot. He said Inferentia would be a cost-effective processor with high throughput, low latency and continuous performance. Inferentia supports popular frameworks such as INT8, FP16 and hybrid precision. More importantly, it supports a variety of machine learning frameworks, including TensorFlow, Caffe2 and ONNX.


Of course, as a product of Amazon, it also supports data from popular AWS products, such as EC2, SageMaker and the new flexible reasoning engine. However, the chip will be officially sold in 2019.

The move is seen as a challenge from AWS to Nvidia and Google


At present, there are so many TensorFlows running on AWS (figure below).


AWS flexible reasoning reduces the cost of in-depth learning by about 75%.


At the meeting, Andy Jassy announced the launch of Amazon Elastic Inference, a new service that allows customers to accelerate GPU-driven reasoning to connect to any Amazon EC2 instance and reduce the cost of in-depth learning by up to 75%. Lei Feng learned that the average utilization rate of GPU for P3 instances is about 10% to 30%, which is a waste of flexible reasoning.


Amazon Elastic Inference will be a very important industry changer, capable of reasoning in a more cost-effective way, suitable for Amazon SageMaker notebook examples and terminals, and supporting machine learning frameworks TensorFlow, Apache MXNet, according to Andy Jassy. And ONNX.

Official Start Block Chain Service: QLDB and Managed Blockchain


Although Andy Jassy made it clear a year ago that AWS was not interested in block chains, today he is very excited to announce that AWS has officially launched two block chain services, one is QLDB (Quantum Ledger database) and the other is AWS Managed Blockchain (management block chain).


When users use QLDB, they do not need to participate in complex development such as taxonomy billing applications, data change history is immutable, and encryption technology is used. Lei noted that QLDB also noted that QLDB is serverless, so it automatically expands to support application requirements, meaning there are no servers to manage and no read or write restrictions to configure.


AWS Managed Blockchain, as a managed block chain service, supports ethernet and Hyperledger Fabric.Andy Jassy, and AWS prefers Hyperledger Fabric

AWS will sell hardware to hybrid cloud customers to run in their own data centers

One of the highlights of today's meeting is the cooperation between AWS and VMware. Pat Gelsinger, CEO of VMware, has also come to Andy Jassy's stage.


On the right is Pat Gelsinger, CEO of VMware.

They jointly announced that VMware would formally deploy AWS's cloud service, and that AWS would allow customers to order the same hardware as their cloud services from next year to run in their own data centers via a service called AWS Outposts. The service will also use software designed by VMware to help integrate the two operating environments. It will go public in 2019. (Lei Feng note: VMware is a leading provider of global desktop to data center virtualization solutions, which can be called


AWS Outposts

However, many details surrounding AWS Outposts remain somewhat unclear, and Andy Jassy did not reveal much about the interview, such as exactly what hardware configuration was provided.

Deep Racer, the first global autonomous racing car with micro-machine learning capability


Andy Jassy is also a sports enthusiast in private. At the scene, he announced a global autonomous racing program called AWS Deep Racer. In competitions around the world, developers will be able to compete with others on the physical track. Amazon will host the AWS Deep Racer Final and present the AWS Deep Racer Championship at next year's re: invent conference.


DeepRacer is a 1 / 18 radio-controlled autonomous four-wheel racing car designed to help developers understand RL reinforcement learning

Launch SageMaker Ground Truth Service: Automatic Labeling


Today, the label task is still determined by users, but AWS announced that SageMaker Ground Truth, a training set label service, is being launched. With Ground Truth, developers can point services to the bucket where data is stored and allow services to automatically tag it. Users can set trust levels for fully automated services, or send data to people. Just as you hate an employee, you can have Ground Truth tagged.

Others, such as AI/ML product matrix, hybrid cloud family and so on, are also published, which will not be discussed here, directly above.



Lei Feng net summary

A CEO has accomplished an excellent product promotion and brand promotion in a technical way.


Andy Jassy's last PPT in 180 minutes

For now, according to Gartner, the global public cloud market will grow to $278 billion in 2021, up from $176 billion this year. In 2022, sales of AWS will reach $71 billion, which will bring AWS's valuation to about $350 billion. Microsoft, Google and Aliyun are catching up. AWS is trying to keep the gap between 2 and 5 years, not only in terms of income, but also in terms of technology and insight.

Andy Jassy felt that whether Google or Microsoft chose to try to replicate the size of AWS in a short period of time, the effect would certainly be limited

The consistent formula within AWS's success is to focus on the needs of customers and to place their views on the world on Web services that meet those needs. With strong management discipline, decentralized teams and data-driven business planning and auditing, AWS's managerial operations and innovative ways have proved to be a competitive advantage.

In the near future, AWS will assume that a new generation of developers will not consider instances, servers, and clusters, but will focus on writing software or possibly buying Lambda functions that will automatically be available and connected to every imaginable area of the infrastructure.

This is what AWS will do next.

Just like that.

But what is certain is that the cloud will not only stay at today's level, it will certainly promote the birth of new products, services and business models.

China IT News APP

Download China IT News APP

Please rate this news

The average score will be displayed after you score.

Post comment

Do not see clearly? Click for a new code.

User comments