AWS Amazon Web Services Global Infrastructure Layout: Weaving an "Invisible Super Network" to Support the Digital World"
If you ask who is the most profitable cloud computing giant today, the answer is undoubtedly AWS (Amazon Web Services).
When many people use AWS, they think it is nothing more than a web console. With a few clicks of the mouse, they can open a cloud server (EC2) and build a database. But have you ever thought about what is behind you when you are in China and click to open a server in the East of the United States (Virginia) or Europe (Frankfurt)?
It is not a castle in the air that enables millions of enterprises and institutions around the world to complete the transmission and calculation of massive data in milliseconds, but a huge physical network that AWS has spent hundreds of billions of dollars on the surface of the earth in the past two decades.
Today, we break away from those boring official white papers and use the first perspective of big vernacular and real people to deeply disassemble the global infrastructure layout of AWS. See how it turns the earth into a "global village" and how its exquisite architecture design ensures that the business is "absolutely downtime.
Popular Disassembly of 1. Core Concepts: Region (Region) and Availability Zone (AZ)
When opening the AWS console for the first time, the dense row of place names in the upper right corner (e. g. US East-Northern Virginia, Asia Pacific-Tokyo, Europe-Ireland) is AWS's
Region (Region)
.
This is the lowest level concept of understanding AWS infrastructure. Many novices, and even some new architects, often confuse "zones" with "availability zones. Let's use a living example:
1. Region-the "city-level" ecosystem
AWS selected "regions" are usually the core economic circles or technology centers of the world. Such as Tokyo, Singapore, Frankfurt.
Each area is completely independent. This means that if the Tokyo area encounters extreme earthquakes or policy changes, the Frankfurt area will not even sneeze. They are completely isolated physically, electrically, and online.
Novice selection principle: the closer to your end user, the better. If your customers are all Japanese netizens, the bucket and server will not choose the Tokyo area because of the short physical distance and the lowest network latency (Latency).
2. Availability Zone (Availability Zone, AZ for short)-the "spare tire cluster" within the zone"
If you think an area refers to a huge computer room, it is a big mistake.
Each region contains at least three or more independent Availability Zones (AZ) ".
What is an Availability Zone?
An availability zone is composed of one or more physically independent "data center clusters" with independent power and cooling systems.
For example, when you deploy your business in the "Tokyo area", AWS actually built three large computer room groups A, B and C in different suburbs of Tokyo (which may be tens of kilometers apart).
Between these three availability zones, using AWS's self-developed ultra-high bandwidth, very low latency
The late optical fibers are dead and connected together. How low is the latency? Usually within 1 to 2 milliseconds. In other words, they are in the same room in the online experience, but in the face of physical disaster, they are independent individuals.
Real-person translation:
The real "high availability architecture" is not that the gambling room does not catch fire, but that your two servers are thrown into zone A and zone B respectively. Even if the available area A is unfortunately hit by heavy rain and power failure, the available area B will take over all traffic instantly, and users will not even be able to perceive the web page turning around.
2. backbone network and edge node: hackers can not break the "private highway"
In addition to building a computer room, another extremely crazy thing AWS has done is
Pull the network cable by yourself
.
Many people think that when we visit servers in the United States in China, the data goes through the ordinary transnational public network (that is, the public network of China Telecom, China Unicom and overseas operators that have changed hands several times). Public network cross-border transmission is not only slow, but also blocked like the morning peak at night.
In order to solve this problem, AWS has laid tens of thousands of kilometers of its own in the depths of the world's oceans and land.
Global Backbone Network
.
This brings us to the third important component of the AWS infrastructure: edge nodes (edge Locations) and
Regional caches (Regional Edge Caches)
.
When you work with AWS's
CloudFront (Content Delivery Network CDN)
Or
Global Accelerator (Global Accelerator)
When used, its workflow is like this:
A user in Paris wants to access your data on a server in Virginia, USA.
Instead of flying across the Atlantic to the United States, requests from Paris users enter AWS's local edge node in Paris.
Once entering the edge node, the traffic is equivalent to getting off the "ordinary public network muddy road" and directly on AWS's private optical fiber highway.
Traffic through AWS's undersea fiber artery, all the way to the source of the United States east.
Through this "two-way public network dash, middle-way intranet long-distance run" method, AWS perfectly bypasses the congestion of public network outlets, and suppresses the delay and packet loss rate of global business to the physical limit.
"Hard Core Black Technology" 3. AWS Infrastructure: Custom Everything (Everything is Self-developed)
Why can AWS play this global network so well? The answer lies in its horror.
self-research ability
. At the size of AWS, there is no ready-made hardware on the market that can meet its needs, so it has started a comprehensive self-research model.
1. Nitro System: Squeeze the "oil and water" of the server dry.
Traditional cloud servers need to take out some CPU computing power to manage virtualization, network and storage. This
Just like in a company, 3 out of 10 employees are in management and only 7 are really working.
AWS is self-developed.
Nitro Chips and Systems
, the management, network, security chores, all solidified into the exclusive hardware chip. Allow 100 of the CPU of the main server to be invested in the user's business calculation, the performance is directly full, and the security is so high that even AWS's own underlying administrators cannot spy on the user's data.
2. Graviton chip: arm wrestling with Intel/AMD
AWS felt that buying someone else's CPU was too expensive and consumed too much power, so it developed it based on ARM architecture.
Graviton Series Processors
. It has now reached the fourth generation. Its characteristic is
Extremely cost-effective and ultra-low energy consumption
. With Graviton examples, enterprises can often directly reduce the calculation cost by 20% to 40%, and properly reduce the cost and increase the efficiency of the artifact.
4. Industry Combat: How to Use AWS's Global Layout for Selection?
After talking about so many hard core details and returning to our actual business, how should we use AWS as a big network to do architecture design? Here are several real-life avoidance and selection experiences in the industry:
Sea short play/game: no brain choice Anycast + backbone network combination if your business is a global game or a cross-border short play of the current fire, users are all over southeast Asia, the Middle East and Europe and the United States. You can put the core server in the "US East" or "Singapore" with the most mature infrastructure, and then put the AWS Global Accelerator on the front end. It uses Anycast (Anycast) technology to allow users around the world to access AWS intranet nearby with the same IP, completely solving the pain points of cross-border McCarton and slow video loading.
Multinational enterprise office: make good use of VPC Peering (peer-to-peer connection) many multinational companies, domestic research and development headquarters, overseas sales division. If each rents a cross-border dedicated line (IEPL), it will cost hundreds of thousands of RMB per month. Smart architects will build VPC in AWS domestic (such as Beijing/Ningxia) and overseas regions respectively, and then open up the intranet through AWS Transit Gateway or VPC Peering through AWS's compliant global backbone network. The security is extremely high, and the cost is charged by traffic, saving a lot of fixed line expenses.
Disaster preparedness is slow: strictly implement "cross-availability zone (Multi-AZ)" deployment. there is a famous saying in AWS circle: "Everything fails, all the time" (anything can break at any time). When buying an EC2 server or RDS database, spend a little more money to check "Multi-AZ (multi-availability zone deployment)". AWS will automatically help you synchronize your data. Once the primary zone is accidentally dropped, the standby zone is seamless in seconds.
Take over, this is the ultimate card to keep the year-end bonus and the company's business.
5. Conclusion: "Roman Avenue" in the Digital Age"
The core reason why ancient Rome was able to rule a huge empire was that it built a hard-core transportation network with "all roads leading to Rome.
Today's AWS is actually playing the role of the road builder in the digital world. It encapsulates the complex global network topology, cross-border submarine optical cables, compliance restrictions on both sides of the ocean, and the physical operation and maintenance of tens of thousands of cabinets into an "out-of-the-box" cloud service.
As developers or business leaders, we don't need to care about the physical details of the bottom layer and how the fiber is spliced. Standing on the shoulders of AWS, your code and business can reach every corner of the world with Internet connectivity in an instant-it is cloud computing that gives the most charming charm of this era.
