Which node does Amazon cloud server choose to go to sea? AWS Japan/South Korea/Singapore/US Computer Room Network Latency Competition
When your business is ready to go abroad and start providing services to overseas users, the first technical hurdle you face is
Select data center (node)
.
Cloud computing boss Amazon cloud (AWS) in the world is full of dense data centers. If you go to the AWS console, Asia Pacific and North America alone, there are a bunch of names such as Tokyo, Seoul, Singapore, Oregon, Virginia, etc.
Many team leaders who have just gone to sea clap their heads: "If there are many Americans, then buy America!" Or "choose Japan for those near us!" As a result, it was discovered after the launch that due to the wrong node selection, the delay of the cross-border network was unbearable, the core business was directly stuck, and the users who bought from overseas ran out instantly.
The node is well chosen, and there is no worry about going to sea. Today's tutorial does not talk about empty official PPT terms, but directly goes to the hardest core.
Full Network Delay Measured Competition
. We will compare the four most popular core nodes in the sea horizontally:
Japan (Tokyo), Korea (Seoul), Singapore, USA (US-West/US-East)
, teach you how to use the vernacular and the clearest business logic, choose the most suitable for your business "the right man".
Core premise: How does network latency occur?
Before entering the red and blue pk of major computer rooms, we must first recognize the cruel iron law of the physical world:
Delay cannot be eliminated, it can only be shortened.
When network data is transmitted across borders, it is buried deep in the Pacific Ocean.
transnational optical cable
. The transmission speed of light in an optical fiber is about 200000 kilometers per second. Packets start from your domestic server or overseas users' mobile phones. Every time they pass through a router and run an extra thousand kilometers, the meter (delay) will pop up.
Within 50ms: Speed. The naked eye is completely non-sense, suitable for MOBA, FPS games, precision quantitative trading.
50ms - 100ms: Smooth. Brushing web pages and App interaction can't feel Caton at all.
100ms - 200ms: Slightly delayed. Watching videos and praising can feel a slight pause and can barely accept it.
More than 200ms: Caton is serious. The web page loads in circles and faces loss of users at any time.
After understanding this passing line, we will take stock of the four hot nodes one by one.
Competitors Dismantling and Real Delay Competition
Let's assume that your core R & D team or initial source station is in mainland China (such as coastal, southern, and eastern China), and see what the real latency data is when overseas users connect to these four nodes.
Player 1: AWS Japan Tokyo Node (ap-northeast-1)-Asia Pacific All-round Six Bailang
Tokyo is the earliest, most mature and most bandwidth home for AWS in Asia.
Domestic connection delay: extremely excellent. If you ping AWS Tokyo from Shanghai, Beijing, or Guangzhou, the latency is usually between 35ms and 60ms. This is even faster than some inter-provincial visits in China.
Delayed overseas users
North American users connected to Tokyo: about 110ms - 130ms (trans-Pacific). Southeast Asian users connected to Tokyo: about 70ms-90ms.
Applicable scenarios: game to sea, cross-border e-commerce, Asia Pacific headquarters. If your main target users are Japan, Taiwan, or need to take into account the high-frequency remote operation and maintenance of domestic research and development teams, the Tokyo node is the golden node in the Asia-Pacific region that can be selected with your eyes closed without making mistakes.
Player 2: AWS South Korea Seoul Node (ap-northeast-2)-"Physical Pin-up" for North China Users"
The physical proximity of the Seoul node to northern China (Shandong, Beijing, Liaoning) is outrageous.
Domestic connection delay: Northern ceiling. If you connect directly to Seoul from Qingdao and Beijing, the delay can be as low as 25ms - 40ms! However, if it is South China (Guangzhou/Shenzhen) in the past, the delay may deteriorate to more than 60ms due to routing detours.
Delay for overseas users: The external radiation capability is slightly lower than that of Tokyo, and the delay for European and American users is usually about 10ms higher than that of Tokyo.
Applicable scene: precision fixed-point strike. If your sea business is only aimed at South Korea, Japan, Hong Kong, Macao and Taiwan, or if your R & D team is concentrated in the north, you can get an excellent development experience by choosing Seoul.
Contestant 3: AWS Singapore Node (ap-southeast-1)-Radiation Southeast Asia's "Sea Gate"
With the large-scale entry of Chinese companies into Southeast Asia (Indonesia, Vietnam, Thailand, Malaysia) in recent years, the Singapore node has become the hottest spot in the whole network.
Domestic connection delay: performance is fair. South China (Guangzhou/Shenzhen) used to be about 40ms-50ms. But if you are in the north (Beijing), the delay will rise to 80ms-100ms.
Overseas user delay: Southeast Asia bully. The entire Indonesia, Thailand, Vietnam users to visit Singapore, the delay can be dead pressure within 15ms-40ms. At the same time, Singapore's delay to India is also very beautiful (about 40ms).
Applicable scenarios: the only solution to the sea in Southeast Asia. As long as your business is to dig for gold in Southeast Asian markets, or to take into account India and the Middle East, don't worry, Singapore is the only destination.
Contestant 4: AWS US Node (US West-Oregon/California, US East-Virginia)-Embrace Global "Super Base Camp"
The American nodes are usually divided into the US-West (Oregon/Northern California on the Pacific side) and the US-East (Virginia on the Atlantic side).
The first choice for the sea team is American West!
Domestic connection delay: physical hard injury. Even AWS West Oregon, which is closest to China, used to need 120ms - 150ms (across the entire Pacific Ocean). If it is Eastern Virginia, the delay will take off directly to 200ms-250ms.
Overseas user delay: localization invincible. The delay of visiting US-West nodes in the United States is 10ms-
Between 60ms. At the same time, the performance of the US-West node radiating to South America (Brazil, Argentina) and Europe is also qualified.
Applicable scenarios: Common platforms for North America, South America and the world.
Ultimate Selection Decision: Teach You to Take Your Seat
After reading so much data, you may still be struggling. We make decisions directly using the crudest "business matching method:
1. If your business is the old problem of "global service:
Best option: Choose AWS Tokyo, Japan.
Why: Tokyo is in the global fiber optic cable transit station. It is fast to the mainland of China, not slow to the United States and West (120ms), and excellent to Southeast Asia. Taking Tokyo as the global base camp can maximize the reconciliation of the delayed contradictions in all corners of the world.
2. If your business is staring at the "Europe, America/North America" market:
Best option: AWS US-west-2, Oregon.
Why: Virginia is too far away. Domestic research and development links it like PPT. In western Oregon, the US users visited quickly, and the 130ms delay during maintenance by the domestic team was completely within the acceptable range.
3. If your business wants to eat the "overseas/Southeast Asia" dividend:
Best option: Choose AWS Singapore.
Why: Southeast Asia's local infrastructure network construction is uneven, so don't put servers in the United States and Europe to allow Southeast Asian users to visit across countries. Deploying the application nearby in Singapore is the only magic weapon to surrender various metaphysical weak network environments in Southeast Asia.
Two advanced self-defense skills (to prevent the upper line from stepping on the pit)
If you have already selected the node, the technical team must do these two things on the architecture before starting up, otherwise the delay will still be off the charts:
Make good use of AWS Global Accelerator (Global Application Accelerator): If your server is bought in the United States and the domestic team or some Asian users are too slow, you can turn on this function in AWS. It will assign you a Anycast static IP. When users connect, they will first enter the local edge computer room of AWS, and then fly all the way through AWS's multinational enterprise intranet line. The delay and packet loss rate will drop precipitously.
Configure CloudFront overseas CDN (static and dynamic separation is required): all pictures, CSS and JS scripts in the web page are hosted on AWS's CloudFront CDN. In this way, no matter which corner of the earth your source station server is in, when American users, Japanese users and European users refresh their web pages, they will open pictures in seconds at the edge nodes at their doorstep. The source station only needs to handle the lightest dynamic interface.
Summary
In the selection logic of Amazon Cloud, there is no "best" node, only "most suitable" node. Summed up into four sentences:
Research and development and Asia-Pacific both choose Tokyo, fixed-point Korea and Japan on Seoul.
Find out the physical location of your core user base and compare the delay red line in this article.
, your first step to the sea has gone steady and accurate.
