Latency and Geography: Why Distance Matters in Site Selection
Light is fast but not instant. That simple fact shapes where certain data centers belong.
Key takeaways
- Light in glass travels at roughly two thirds of its speed in a vacuum, so distance always adds delay.
- Real fiber routes are longer than straight lines, and equipment adds delay on top.
- Latency sensitive workloads belong close to users. Many others can sit far away where land and power are easier.
People often say the internet is instant. It is not. Every signal takes time to travel, and that time grows with distance. For some data center uses this matters a great deal. For others it hardly matters at all.
The physics in plain terms
Fiber carries data as light. Light moves fastest in a vacuum. Inside glass it slows down, to roughly two thirds of that top speed. As a rule of thumb, that works out to about one millisecond of one way delay for every 200 kilometers of fiber, or about five microseconds per kilometer. A millisecond is one thousandth of a second.
This is a physical limit. Better equipment cannot remove it. Only a shorter path can.
Why the real number is higher
- Routes are not straight: cable follows roads, railways and rights of way, so the fiber path is longer than the map distance.
- Equipment adds time: every router, switch and amplifier along the way adds a small delay.
- Round trips count: most services need a request and a reply, so delay is felt twice.
- Congestion: busy links can add variable delay.
Which workloads care
Some applications are sensitive to delay. Examples include online gaming, voice and video calls, financial trading, and interactive control systems. Others are tolerant. Large data storage, backup, batch analysis and a lot of AI model training can run far from users because they do not need instant response.
AI deserves a specific note. Training a model is usually tolerant of distance from end users, while serving answers to people is more sensitive. These can be placed in different locations.
How this changes site choice
Delay-sensitive services pull data centers toward large population centers and exchange points. Delay-tolerant services can move to places with cheaper land, more available power or cooler climates. This is one reason some regions grow dense clusters while others host very large, remote campuses.
The tradeoff is real. A remote site may offer power and space. A metro site offers proximity but is often limited on both.
Questions to ask
- Who are the users, and where are they?
- What response time does the application actually need?
- What is the measured fiber route distance, not the straight line?
- Can the workload be split, with delay-sensitive parts near users and heavy processing elsewhere?
- Are there multiple routes with similar delay, so failover does not slow service sharply?
The lesson is simple. Distance is a cost, and the right amount of it depends on the job.
A worked example in words
Suppose two cities are 1,000 kilometers apart on a map. A fiber route between them may be 1,300 kilometers or more. Using the rule of thumb, one way delay in the glass alone is about 6 to 7 milliseconds, and a round trip is double that, before any equipment is counted. That is fine for a web page and too slow for some trading systems.
This example is illustrative, not a measurement of any real route. Actual numbers should always be taken from tests on the specific path. Providers can often share typical round trip times between points, and independent tools can measure them directly.
This article is general information for public education. It is not legal, tax, financial or engineering advice. Rules and conditions differ by location, so check with qualified local professionals. Spot an error? Write to info@datacenteroutreach.com and we will correct it.