Are replication data-centers useful for speeding up responses?

Question

There are different types of data-centers so I want to explain that I'm talking about data-centers setup with the purpose of providing content faster to a certain local on the planet.

From what I understand, each data-center shares access to the data storage and basically just exists somewhere in the world to keep requests from travailing thousands of miles to the main server.

However, since the each data center has to keep the data in sync with the others wouldn't this slow everything down to negate the benefit of having them in different parts of the world?

Say for example, that a result set had to be fetch from the main server back home - wouldn't that be the same as the user just making the request to that home server himself?

Obviously not, because businesses use them so perhaps someone can explain this.

Answer 1

The short answer is that overall the data transfer is slowed as you expect. But, because the transfer to branch centers can occur on non-peak times (when fewer users are impacted), perceived performance is improved as the user would access the data from a quicker location.

Answer 2

From what I understand there is always a "live" server for each user's data, and all requests are fed to the live server. Synchronization occurs in off-peak moments. When a user moves around, the system will wait until their data is fully synchronized, and then migrate them to the nearest server.

I haven't worked on such a system myself, but that's my understanding of how it works.