Let customers use localized public cloud

Changing Customer Needs: On-Premises Public Cloud

When we talked about the cloud in the past, we must think of a large, centralized public cloud. In the past ten years of development, Alibaba Cloud has also done the same. But since the year before last, Alibaba Cloud has seen more and more new localized deployments appear. This is inseparable from the development of 5G technology and new interactive forms. They put forward new requirements for the position of computing power:

1. Pan-entertainment scene. Live broadcast interaction has penetrated into all aspects of life. The emergence of XR scenarios such as cloud games, AR augmented reality/VR (virtual reality)/MR (mixed reality) has brought new challenges to cloud infrastructure. These scenarios have very high requirements for delay. high.

2. Industrial parks. How the industrial park connects to the cloud, or how to customize a cloud that suits him, can bring a better experience to the customers he serves.

3. Self-built IDC to migrate to the cloud. Some customers have their own offline IDC data centers and original technology stacks. They see the value of the cloud, but the self-built IDC is far away from the public cloud center Region, and there is a delay in business active-active and disaster recovery. and cost issues.

4. Local specific enterprises. Some specific industries, such as government and enterprise customers, may have some requirements for local processing of data and not going out of provinces.

In this regard, Alibaba Cloud has been thinking: as a public cloud, how to get closer to customers and better assist customers in business innovation. We mainly made the following analysis on the whole thinking path:

First of all, from the perspective of customers, in the past, customers only need to deploy services in first-tier cities such as Beijing, Shanghai, Hangzhou, and Shenzhen. At present, in order to process a large amount of data or meet the requirements of low latency, localized deployment has become a very clear appeal of customers.

Taking Beijing Region as an example, if the customer's business system is also in Beijing, the latency is very low. However, if the enterprise is located in Wuhan or Xi'an, the delay to Beijing may be relatively high. Alibaba Cloud has already covered a wide range of first-tier cities, and we have seen some more extensive second-tier provincial capital cities, as well as third- and fourth-tier cities' deployment demands, including edge nodes.

The second point is that today, no one doubts the value of the cloud to business innovation. Therefore, customers will hope to enjoy the public cloud experience, including a consistent account system, console, operation and maintenance, cloud network, etc.

Four deployment forms of Feitian architecture

In the speech at the main forum, Alibaba Cloud released a cloud multi-form deployment architecture. It can be roughly divided into four deployment forms:

First of all, the central region, which mainly covers first-tier cities, provides the strongest computing power, the most abundant product capabilities, and elastic capabilities, etc., to support various types of extensive innovation.

Second, the local Region service officially released today is mainly for some regions with relatively developed digital economy and relatively active demand for IT. Based on the same Feitian architecture, we mainly export the five product lines of the most basic computing, storage, network, security, and database to bring customers a low-latency, low-cost, and consistent experience with public clouds.

The third is to go further down, Alibaba Cloud has edge cloud nodes. The last one is the on-site compute node.

**Back to the architecture point of view, how did Alibaba Cloud get to the present step by step?

**
(Below) The far left shows the earliest central region of Alibaba Cloud. All product capabilities are designed based on the central region. This is a cloud operating system built to manage tens of thousands or hundreds of thousands of servers. Such a large and comprehensive capability will encounter more challenges when outputting close to customers.

In the process of sinking, the demand and scale of resource deployment in some areas are not so large, and only a few hundred servers may be needed. The local Region has adjusted this, but all designs still maintain a unified architecture; therefore, a consistent product experience can be maintained, including OpenAPI, tool chain, deployment, operation and maintenance, etc., are all exactly the same. When using the product, you have no idea whether the product is deployed in a central city or a second- and third-tier city.

Edge nodes will be smaller in size. Based on the same architecture, Alibaba Cloud has launched edge zone products and cloud boxes. The cloud box is oriented to the interior of the enterprise IDC, and is output to the enterprise's existing data center in the form of a cabinet.

Local Region: a deployment method for low latency and strong interaction
Among the above four deployment forms, first introduce the local Region service, which is applicable to the aforementioned scenarios such as pan-entertainment, customer IDC migration to the cloud, industrial parks, local industries, and government and enterprise customers going to the cloud.

At the bottom layer, customers can access Alibaba Cloud's local Region through the cloud network to complete nearby data processing and low-latency access. When there is further large-scale elastic interaction or cross-region interaction, it can be upgraded to the central region to complete unified resource deployment, scheduling and access. Such a structure can help enterprises allocate resources more reasonably.

Taking the live interactive scene as an example, we list a reference architecture:

Today's users or anchors are not all in first-tier cities, but may be in second-tier or third-tier cities. The live broadcast platform can access the local Region nearby through the public network IP or the customer's physical dedicated line to achieve low-latency access, and use the local Region to complete video and front-end business processing. Most customers carry out national or even global businesses, and further upward convergence can use the central Region to achieve global processing. The data synchronization between the center and the local can also be completed conveniently, reducing and avoiding the problems caused by the customer's original data out of synchronization between IDCs.

This set of architecture is based on unified product capabilities, unified operation and maintenance standards, and unified API. All development, operation and maintenance, and online deployment of customers are very convenient.

Cloud box: public cloud deployed in the client computer room

The central Region and the local Region are still relatively large in scale and volume. At present, many customer innovations require data to be processed locally, such as:

For video surveillance on highways and smart communities, the data needs to reside locally. If a large amount of video data is directly uploaded to the cloud, the network cost is still relatively high.
During the digital transformation of manufacturing enterprises, a large amount of data will be generated and applied to the production system, which has strict requirements on delay.

Multi-regional branch offices have the need for unified operation and maintenance.

Cloud box products can well meet the above scenario requirements, provide the same experience and automatic operation and maintenance capabilities as public clouds, and users can enjoy the latest and highest quality cloud products and services at any time.


As a public cloud, all our products are online. The first step in using the cloud box is to place an order on the official website. You only need to select and match your resources according to the SKU, and all your expenses will be generated online in the end. After you place an order, we will finally complete all the pre-installation of software and hardware integration in the factory, and transport it to the customer's computer room. After the power is connected to the network, the user can see the product instance on the public cloud console, and its use is the same as that of the public cloud. There are standardized processes for daily operation and maintenance, hardware replacement, and removal after expiration, minimizing your daily IT operation and maintenance work.

Let's look at the picture below. The gray in the upper right corner of the picture is the original enterprise equipment, and the orange in the lower right corner is the cloud box. On the far left are the customer's public cloud resources.

If you don’t look at the orange cloud box device in the lower right corner, in the past, if you wanted to use resources on the cloud, you needed to use a dedicated line or access the public cloud through the cloud network. The maintenance of offline IT facilities involves procurement of hardware, system maintenance, patch upgrades, fault replacement, etc., which is quite cumbersome, while the cloud box is managed by a unified operation and maintenance platform.

The cloud box can communicate with the customer's local equipment. In the local network, you can monitor all traffic and set corresponding security protection policies. At the same time, when there is an elastic and sudden demand, you can use public cloud resources at any time. The public cloud uses a unified private network VPC, which is very convenient to manage.

Cloud Box can not only meet customer data local processing requirements, but also provide the latest product capabilities to the public cloud.

Summarized by this radar chart, Alibaba Cloud first started to build a public cloud from the central Region. In the past, it could only satisfy customers' services for or relying on central cities, but it was difficult to meet the large demand for sinking computing power.

Today, customer needs are everywhere. Through the evolution of the entire architecture, the four deployment forms of Feitian Yiduo Cloud can meet the requirements of customers in different locations for delay, computing power, compliance, etc., and provide a consistent product experience to help Customer business innovation.