It is crystal little is understood that after the first working day National Day holiday, AWS released in China on the region a "heavyweight" container service called AWS Fargate, halo new operating AWS China (Beijing) and the Western cloud data AWS China (Ningxia) regional operators are to provide the service.
It is reported that this is an Amazon ECS can be applied to the calculation engine, focused on helping enterprises run container in the production process, but without having to deploy or manage the server, in other words, focus on the design and build applications without having to keep in mind too much infrastructure "that thing."
Have to say, the demand for container services positioning, AWS Fargate can be described as a hit in, this application modernization products from AWS market leader introduced Aaron Kao's we can deeply feel.
For some time, led to the AWS cloud service provider you will hear a lot of similar user feedback, in which to build their own applications become a modern enterprise had to actively pursue and achieve. What is the modern application? Globalization must, mass, millisecond delay, PB-level data and multi-modal, and many other properties are essential.
Back in 2000, Amazon's application to build the nature of the monomer (monolithic application), very large but the lack of agility performance. In particular, if you want to add new products, we must do the source code editing and modification, many teams need to focus on development applications in a single pipe, low efficiency can be imagined.
"In 2001, we want to change that way to build applications, so it is broken down into multiple micro-architecture services, to create 'two-pizza team', which is a change in the organization, it is to reach a more rapid development efficiency. "talk, the less that crystal, the so-called" two-pizza teams on "intuitive sense refers to the number of team members only need to meet two pizzas appetite; it extended this configuration is solely responsible for its own development team of service on the application of absolute ownership, from discovery application, and other aspects of development to deployment, with full accountability mechanisms, have claimed that this is the DevOps model development and application.
From transition to a single micro-services, we also found that calculate changes over time, from the virtual machine to the various instances of the type, then the rise of container, non-server computing rise of Lambda. We see a lot of customers to do a lot of interesting things in terms of container, such as McDonald's, we launched a home delivery service, and is reached using a micro-services running on Amazon ECS within four months.
It is reported that this cloud of native micro-service architecture can be less than 100 ms of delay spread twenty thousand orders per second, and can easily and integrated global delivery partners. This is the application of modern exploration of the Amazon build mode, and now thousands of users have already embarked on this adventure.
Back to the AWS Fargate freshly baked, in the final analysis or the use of container inspired to bring in the process of building a modern user applications. In this process, we often think of a question, how to integrate the container on the opponent? After using Fargate service, users do not need to personally deal with server and infrastructure aspects of the work, can do rapid introduction and expansion of container applications, this process can only be done under the circumstances to pay to use their services.
No mention secure server experience, without having to spend too much energy requirements management infrastructure, we still remember, in 2014, re: Invent conference published Lambda, Lambda month and now hundreds of thousands of tens of thousands of active clients million times executed. As a non-event-driven application server computing, Lambda allow users with or without pre-management server to run code, simply pay by computing time is consumed.
Specifically, it can be automatically extended to the workload for popular IDE kit, can also be improved by simplifying the application deployment nesting efficiency; Further increasing integration function, such as an integrated Lambda application load balancing, and the like Lambda SQS .
Take, for example, as a user of using AWS Lambda, Lambda efficient use of Coca-Cola and STEP function to update the vending permit applications, and to facilitate the development of innovative personnel calculation example by Lambda management, and even can be done so that engineers in rapidly build and deploy new capabilities within a few days. Up to now the incomplete statistics, hundreds of thousands of users have successfully established micro-services and applications in the process of modernization in the use of AWS Lambda.
The expression here, and the relationship between Fargate Lambda, there may be a lot of people feel confused, is also a focus on applications and upper layer architecture, the difference between them? Essentially, Fargate and Lambda are two different ways to build customer applications and packaging. If you want to start running container level, it becomes the type of container and packaging services, that can be used Fargate; if you just want to pack the code as a function, then you can use the trigger mention Lambda, simply, users want how much, in order to measure the degree of abstraction involved in control.
For example, FINRA FINRA, which is mainly for tracking thousands of the US Securities and Exchange Stock market traders, billions of dollars in transactions, detect possible fraud. "This job content decisions sometimes encounter peak, the inadequate use of resources now, so you can switch to the Lambda calculation speed to reach and satisfy peak demand; well-known US media company Turner, because do not want to spend too much energy in charge of its many instances, it will be migrated to Fargate, started to build a website sharing service platform, not only to reduce costs but also shortens development time. "
In this way, from product and service trigger, we observed that the overall pattern of AWS container services, universality is a outstanding features. "Overall covers three levels, the bottom is elastic image registry, namely Amazon ECR (Amazon Elastic Container Registry), image files can be stored in a container and very easy to use; the middle layer is a calculation engine to run in a container computing platform on you can use Amazon EC2 as the startup type, you can use AWS Fargate to start without a server running and so on; moreover orchestration layer can be selected according to customer demand, if not own a clear tendency, if required abstraction level low and focus on basic architecture or configuration, users can also build their own choreography service. "Aaron Kao expressed.
After the talk as a whole, the focus now fire K8S, Aaron Kao further elaborated over time, users have mentioned a lot of demand on Kubernetes, perhaps this is the reason to launch Amazon Elastic Kubernetes Services, and that the AWS Kubernetes exactly what are the advantages?
As we all know, EKS is run production-level workloads platform, safe and secure is a top priority. "We offer original and upstream (native and upstream) of Kubernetes experience, Kubernetes run of Kubernetes AWS services and open source exactly the same level; in addition, we also work together with the community, will contribute code to the upstream."
Needless to say, the user selects the beginning of AWS is very important that the depth and breadth of services, 165 kinds of different types of service from the cloud storage services, management services, can effectively help in the EKS achieving seamless integration with AWS platform and comprehensive release capability can not be underestimated.
Specifically, the user how to use Amazon EKS it? The whole process including through micro-service, platform as a service run this architecture, of course, the use of EKS migration to enterprise applications, some users also involves the use of EKS for machine learning. "In the future we have to do something to ensure EKS for any customer who wants to use the AWS are extremely friendly, such as making EKS with AWS Identity and Access Management (IAM) integration, each of EKS Pods are available IAM license believe EKS soon AWS in China (Beijing) region, AWS China (Ningxia) region, AWS South America (Sao Paulo) region launched, worth mentioning that currently 51% of Kubernetes workloads are running on AWS. "
EKS use in addition to more than a friendly, KAO represent the more important innovations will be concentrated in the direction of the whole stack, such as hardware-level Nitro system, improving performance and hardware security, and let Hypervisor (hypervisor) more operate effectively; software level Firecracker, as an open source micro virtual machine, allowing users to run when the vessel can be more secure.
It is crystal less informed news for this new virtualization technologies and open source projects run multi-tenant workloads container that allows service owners will realize the speed of the vessel, resource efficiency and performance with traditional virtual machine security and isolation binding, thereby running container-based, secure multi-tenant service.
Firecracker Run hypervisor (VMM) on (KVM), and provides a RESTful API to create and manage vCPU and memory with any combination of micro virtual machine in Linux kernel-based virtual machine, to meet the needs of the application.
As an open source project under the Apache 2.0, low analog device characteristics can shorten the startup time and reduce memory footprint of each micro virtual machine and provides a reliable container for each sandbox environment. Firecracker currently supports Intel processors, and will introduce support for AMD and ARM in 2019.
"We work closely with our customers, their closeness can focus on performance in 90% of the roadmap are derived from the user, they will tell us exactly what, hope to reach the technology in the future will be how? Our future focus is to focus on security , availability, reliability, and innovative developer tools, but user demand is the key, it is absolutely important. "
