Shulou(Shulou.com)06/03 Report--

The annual AWS Re:Invent conference, one of the technological and commercial bellwethers in cloud computing, was held last week in Russ, USA. As in previous sessions, AWS has released hundreds of new products or technologies intensively. With the increasing popularity of cloud computing, especially public cloud in China in the past two years, domestic technical media and developers pay close attention to AWS Re:Invent, and a lot of information is transmitted to China almost without time difference with the United States. The previous container Rubik's cube's technical article impressed everyone, and many fans asked us to talk about our understanding of the AWS Re:Invent conference from our own perspective, so we gave a complete interpretation as soon as the conference was over.

Among the new products released by AWS every year, the areas that attract the most attention are the traditional cloud computing infrastructure and databases, which are currently ahead of all AWS's competitors.

The released EC2 A1 instance of ARM architecture, EC2 C5n instance of 100G network, new file storage service supporting Windows NTFS and Lustre protocols, DataSync synchronized with cloud data, ultra-low-cost cold storage S3 Glacier Deep Archive, global accelerated network Global Accelerator,Serverless database Aurora and DynanoDB based on intelligent routing, new timing database Timestream, blockchain database QLDB, etc. All show the strong strength of AWS in infrastructure layer and basic data middleware. Especially for servers with ARM architecture, since Samsung, Nvidia and Qualcomm gradually abandoned this field, AWS announced that the launch of ARM architecture instances has played a significant role in promoting the ecology of this field.

In the field of AI, AWS has also responded to a series of AI initiatives by Microsoft Azure and Google Cloud this year, releasing a large number of AI applications and AI infrastructure services, such as dedicated Inferentia chips for reasoning scenarios and elastic computing Elastic Inference based on elastic GPU resource pool, expanding SageMaker into data marking service Ground Truth and reinforcement learning SageMaker RL There are three sets of machine learning domain services: AmazonTextract of OCR scenario, Amazon Personalize of recommendation system, Amazon Forecast of prediction system, RoboMaker of robot application development service, and Marketplace for machine learning algorithms. During the conference, AWS creatively held an autopilot competition:

DeepRacer League, that is, for the attendees, design the AI algorithm on the DeepRacer × × which is the size of the real car 1ax 18 provided by AWS to win the race. Reminds me of the famous Sony robot dog of the RoboCup Robot Soccer World Cup. However, AWS does not have a leading edge in the AI field as a whole. Relatively speaking, Google has stronger technical strength than AWS in the AI infrastructure layer and Microsoft in the AI application layer.

But the rich customer scenarios accumulated by AWS in the public cloud market over the years, as well as the full stack of technology from hardware to software, support AWS to have a stronger possibility of business innovation, with SageMaker released last year as a typical example. The Inference two-piece suite released this time is an important service launched by AWS in the field of AI based on its strong infrastructure layer capabilities, which is worthy of continued attention. By the way, Huawei HC conference in October also launched the AI dedicated chip Teng Asend, which is now in mass production and use in Huawei Cloud EI products.

Then there is the Serverless Computing area that AWS has always valued very much, that is, Lambda.

At present, of all the public cloud players, I think only AWS is very optimistic and committed to this field, while other players are more likely to follow. It's been four years since AWS launched Lambda in 2014, and if you follow this service, you may notice that Lambda is far ahead in market statistics every year, and all kinds of open source projects around Function emerge one after another. But these figures have little real business value, because the entire Function market is pitifully small. For example, in AWS's official Serverless application market, there are very few applications, and the top ones are all Amazon's own services, such as a gadget written for Amazon Alexa at the top, and even this deployment is only 20, 000, which is negligible compared to more mainstream container or virtual machine applications.

However, the AWS conference carefully prepared a series of releases for Lambda, from the open source Lambda sandbox technology Firecracker, the release Lambda Custom Runtime announced that Lambda will be able to support any development language, to provide a series of tool chains to support the development of Lambda applications For example, the Lambda plug-ins of various mainstream IDE, the component tool Lambda Layers, the complex program tool Nested Applications, the growing integration of various middleware services, and even an architecture self-inspection tool Well-Architected Tool. All releases are released in AWS CTO Werner Vogels's Keynote, which shows how much AWS attaches importance to it.

These releases of Lambda represent a new round of thinking about Serverless Computing by AWS. Open source Firecracker and supporting Custom Runtime are actually a kind of defensive behavior in response to the increasingly hot view in the industry that the container technology represented by Serverless Container is the real Serverless Computing. The language relevance of Lambda and the non-containerized sandboxing technology have always been questioned by the outside world, so a large number of open source projects focus on using containers to build Function, such as Kubeless and Knative in K8S ecology. It is expected that the Lambda successor will continue to integrate with the container ecology, such as Firecracker as a K8S container runtime choice.

And a series of Lambda development tool chains are AWS's efforts to promote the "functionalization" of business logic from the perspective of development, to solve the biggest pain point in the Lambda market: unable to write real production applications that can really solve business problems. AWS CEO mentioned in the conference interview that AWS believes that future edge computing is one of the application scenarios of hybrid cloud, and that edge and Internet of things scenarios are one of the main battlefields that Lambda can play in the future.

Lambda's easy development, event-driven, on-demand startup and shutdown, and low resource consumption are indeed very suitable for edge computing, while AWS's edge solution has been promoting the use of Lambda to connect cloud services, while Microsoft and Google are the main push to move cloud services, especially AI. However, whether Serverless Computing, which Lambda represents, will become a mainstream computing model in edge computing and Internet of things scenarios in the future still needs more customer scenarios to verify.

Finally, there is the field of cloud native computing represented by containers and microservices, which is in fact an area where AWS has always been weak.

On the one hand, almost all the technological innovation in this field comes from its competitors Google and Microsoft, on the other hand, AWS does not want to promote the development of the container market because of its long-term traditional market advantages. So we saw Fargate that messed up last year, which means that AWS wants to weaken the contrast between containers and virtual machines, but emphasizes the application experience.

This time, AWS still emphasizes that it is the most popular public cloud for container and K8S, but does not talk about how many customers use AWS EKS or build or install K8S distribution directly on EC2. Of course, the Firecraker released by AWS is also in response to the container runtime technology, especially the gvisor developed by Google, but why Fargate for EKS is still not released one year after it has been announced has to make people suspect that AWS still does not have a clear product strategy for K8S business model. After all, there is a cross between Fargate for EKS and Lambda and EC2. And the Marketplace for Container released this time is just following up on customer demands.

The only interesting release is AWS App Mesh, which is AWS's follow-up to Service Mesh technology. Oddly enough, AWS uses the Istio community's Istio-Proxy project Envoy, but does not use the more important Istio Control-Plane control surface. The official explanation is that you need to support both ECS and EKS scenarios (after all, Istio can only support K8S but not AWS's own ECS), but without the control plane ecology of Istio, will anyone use AWS App Mesh? We'll see.

AWS has also launched some services in other areas such as security, cloud management and developer tools, such as Control Tower for multi-account management on the cloud, Lake Formation, a data lake construction tool, Security Hub for centralized management of security services, etc., while hot spots in the industry have also followed up this time, such as the launch of blockchain service Managed Blockchain and Managed Kafka services with the increasing popularity of the Kafka project.

The finale is AWS Outposts service, which is AWS's soft and hard all-stack all-in-one solution. After being powered on in the client's computer room, you can build a private cloud directly or form a hybrid cloud with AWS.

This should be the most important launch of the AWS conference, announcing that 10 years later, AWS has finally entered the private cloud and traditional IT market that it has vowed to subvert. Amazon officially listens to customer needs. After all, AWS's corporate philosophy is customer-centric, but it is generally believed that AWS has set its sights on this far larger market than public cloud, because after long-term technology accumulation and acquisition, AWS is no longer just a software company, it has enough hardware technology to enter this market, the only problem is that private cloud requires much more than online services. The sales network and maintenance service of the enterprise market is the core, and it takes many years of operation.

Faced with this question, AWS gave the answer is Vmware, this good partner for many years launched Vmware Cloud on AWS Outposts at the AWS conference, with Vmware's enterprise market, will anyone doubt the determination of AWS Outposts?

Of course, next year's launch of AWS Native is the key, AWS software and hardware full-stack service is the product that AWS can target competitors. In the direction of hybrid cloud, Google and Microsoft chose a pure soft route to cooperate with server manufacturers, while AWS walked out of the soft and hard full stack route, which is really domineering and worthy of continuous attention. By the way, Huawei's hybrid cloud solution is also full of hardware and software. If you are interested, you can follow Huawei ForeCloud Stack (FCS) solution.

I have always had a point of view, both belong to Internet companies, compared to Google pay more attention to technological innovation, Amazon pays more attention to business innovation. The satellite data service Ground Station released by AWS once again gives me this feeling, which has a strong industry understanding and exploration of the future development of public cloud services.

AWS uncovers customer problems from cases where its customers in the aerospace industry use AWS services and creatively develops its extension into a public cloud service. This is a very Amazon way of thinking about business. Interested students can check all kinds of online articles at will, so I won't introduce them in detail here.

The above is the complete interpretation of AWS Re:Invent 2018 by Container Cube. If you have different views or would like Container Cube to interpret a product or technology trend for you in detail, please leave a message under the article. The concept of Amazon is customer-centered, the concept of container Rubik's cube is fan-centered, and the number of likes left is more than 200. the boss will kick the editor's ass to bring you in-depth interpretation.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.