1. The "Cloud" Is Really Buildings Full of Servers With Power and Networks
The word "cloud" makes it seem like data floats somewhere in the sky. But if you read what companies explain, the reality is servers (computers) inside buildings called data centers. AWS describes an "Availability Zone" (AZ) as one or more data centers that have separate power and network connections from other AZs. A "Region" is a collection of three or more of these AZs [1]. It is important to know that one AZ is not always just one building. Azure also explains that availability zones are groups of independent data centers with separate power, cooling, and networks [2]. Even though the names differ by company, the idea of not keeping everything in one place is common.
2. Your Data May Be Copied to Multiple Buildings
Whether data is placed in multiple buildings depends on the service and settings. For example, the standard storage class of AWS's S3 service saves the same data on multiple devices across at least three AZs within a Region [4]. This is like copying an important notebook and keeping it at school, your house, and your grandparents' house. However, the company's system does this copying automatically. The designed durability is 99.999999999% over one year, which is a number with eleven 9s [4]. If you subtract this from 100, the loss rate is 0.000000001%, or one in 100 billion. For 10 million stored items, that rate would work out to an average of one lost item every 10,000 years. This is a design value. Some classes store data in only one AZ, so not all services are protected in the same way [4].
3. Why Data Center Zones Are Placed Neither Too Close Nor Too Far Apart
Azure explains that the distance between zones is close enough to connect with low delay, but far enough to separate faults like storms or power outages. When using multiple Regions, longer distances improve recovery from large natural disasters, but they increase delay [2]. AWS explains that AZs are kept more than a few kilometers apart, but all within 100 km (about 62 miles) of each other [1]. It seems zones are placed at a distance that separates faults but does not make communication delays too large. This balance appears in these numbers. The idea that longer distances are stronger against disasters applies when using multiple Regions. However, the design is not uniform. According to Google's English page, Regions usually have three or more zones. But in places like Osaka, three zones may be inside one or two physical data centers, and they are expanding [3]. Because this varies by company and time, it is safer not to assume "everywhere is the same."
4. About 60% of Japan's Data Centers Are in the Tokyo Area
According to a Ministry of Internal Affairs and Communications document (October 2024) on the Cabinet Office website, about 60% of data centers in Japan are in the Kanto region (57% in the Tokyo area alone), and about 20% are in Kansai. By region, Kanto is 61.1%, Kansai is 24.3%, Kyushu is 2.8%, and Hokkaido is 2.0% [5]. Kanto has about 30 times as many as Hokkaido. The document lists latency (communication delay) and transport access as reasons [5]. The number of connections at Internet Exchange points (IX) is 74.2% in Kanto and 24.2% in Kansai. Landing stations for international submarine cables are 53.7% in Kanto and 26.8% in Kansai, so key communication hubs are also concentrated in the same areas [5]. A real estate company, JLL, lists conditions for location: hubs where telecom operators connect are in Tokyo and Osaka, high-voltage power is available, and it is easier to avoid natural disasters like floods [7]. Note that this is a secondary source. On the other hand, many small data centers exist in rural areas for backup and local business needs [5].
5. The Risk of Concentrating Data Centers in One Area, and Ishikari in Hokkaido as an Example
A Ministry of Internal Affairs and Communications plan (April 2023) states that about 60% of data centers are concentrated in the Tokyo area. It warns that if the Tokyo or Osaka areas are hit by a major earthquake, it could greatly affect communication nationwide, so spreading them out is needed [6]. It mentions that many Pacific-side submarine cables were cut during the 2011 Great East Japan Earthquake [6]. The government decided in June 2022 to support building data centers at 7 sites outside the Tokyo area. The policy is to create third and fourth core hubs in Hokkaido and Kyushu, away from Tokyo and Osaka, where renewable energy and submarine cable landing are possible [6]. In Ishikari City, Hokkaido, a data center opened on November 15, 2011. It uses "outside air cooling" to cool server rooms with Hokkaido's cold air. The city's page says this reduces power consumption by about 40% compared to typical urban data centers [8]. The operator also cites that the area is less affected by large earthquakes [9]. The 40% figure is a city introduction number with the condition of comparison to urban types. It cannot be applied to other facilities.
6. Using Two Sites in Tokyo and Osaka, and Who Decides the Location
There are examples of using two sites in Tokyo and Osaka. The AWS Tokyo Region opened on March 2, 2011. The Osaka Region opened ten years later on the same day. Osaka has three AZs. They are about 400 km (about 248 miles) apart in a straight line, so reports said it is easier to build disaster measures with two sites [10]. News reports give the distance as 400 km, but AWS does not confirm that figure on its official pages. In some AWS and Google Cloud services, users can choose which Region to store data in. AWS explains that users choose the Region for their data, and it is not moved or copied outside that Region without consent [11]. Google Cloud also says you can set specific data locations for certain services [3]. However, "choosing" depends on the service, and you must check the terms for exceptions like backup or operational data. There are also government standards. The Digital Agency's standards for municipal systems (October 2022) require cloud providers to place data centers in Japan, keep information assets in Japan unless instructed otherwise, and use domestic sites for backup during faults. They also require independent connections from East Japan and West Japan [12]. This is a standard for buying municipal systems, not a legal obligation for all private services.
7. How to Compare Possible Data Center Locations on a Map
Open a map of Japan and look for Tokyo, Osaka, Ishikari City in Hokkaido, and Kyushu. If you were to place data centers in two places, where would you choose? Think of one advantage of placing them close together, and one advantage of placing them far apart. If you want to look further, ask your parents or teacher for the name of the cloud service you use now. Look for what the company's explanation page says in a language you can read about where data is stored. The wording is different for each company.