A very constant question we receive here at Cloud8 is how to choose the geographic region (set of datacenters in the AWS definition) to place your cloud infrastructure.

There is no formula, but you must take into account at least these three factors:

• costs;
• network latency -> access performance;
• supported features

Costs

Naturally, cost is the first factor that stands out. And the difference between regions is huge. The most affordable regions are USA / East Coast (which is the comparative basis) and USA/West Coast 2.

Note: Costs for all components such as disks, data transfer, and other services have not been included. For full details, visit the AWS costs page .

Latency / Performance

Performance is also equally important. Depending on the performance required by your system and the need to be better ranked by Google (the faster and less latency, you gain points in the search rankings), the best thing to do is to make them as close to your users as possible.

It is difficult to put the regions in an order because we do not know how operators will route IP traffic. But for a website / system that has its main audience in Brazil, an order with a good chance of working on any telco would be:

• Brazil;
• USA / East Coast;
• USA / West Coast 1 and 2;
• Europe;
• 3 regions in Asia.

Supported Features

This item is the least important and tends to become less and less relevant as Amazon is increasingly launching simultaneously in all regions. But it is always important to check whether the architecture you are going to build has all the components in the regions you choose. For example, certain machine profiles only exist in the US/East Coast and until recently services such as DynamoDB and Storage Gateway did not exist in Brazil.

Tips

There are no universal tips that suit all cases. It is always case by case and you must study the cost structure versus performance to arrive at the solution that best suits you.

In the case of Cloud8, we chose to host the site in Brazil due to the performance and network latency that count in Google’s ranking. The main part of the infrastructure management application is in the USA due to costs and because network latency is not important, given that the panel is a rich interface, with high performance and that transfers so little data that it does not depend on where you are. Finally, we spread servers across other regions as Cloud8 internal process nodes.

If you have any tips or comments, feel free to send us a message.