Migration of a large application to Amazon Web Services
The customer had a compute-intensive, internationally used application in a classic data center in operation. The growing complexity of the application has resulted in a steady increase in hosting costs. Due to the static definition of the servers, the power peaks occurring for a few hours a day have determined the operating costs for every 24 hours, even for the less active night hours.
Due to the high strategic importance of the application, the request was to increase operational security and to perform extensions of the infrastructure centrally by the software manufacturer. At the same time, the aim was to increase the degree of redundancy.
In the first step, the application was modularized so that dockerization was possible. Then the entire infrastructure was defined as code and provided with a dynamic hosting metric.
Changes to the infrastructure can now be tracked using a standard software versioning system and are therefore completely transparent. A check according to the four-eyes-principle, such as the test-wise execution of structural changes on an integration environment, is possible.
Since the load is not constant during the day, more servers are started dynamically as users increase. For example, twelve servers are running at noon. By contrast, during the night, the usage is low, so that the number of servers can be reduced to a minimum of two servers. For security reasons, the application is located at two different data center locations. If one datacenter has a problem, one server is still working in the other.
The number of running servers is therefore proportional to the number of (active) users. Peak loads in the morning and late afternoon are cushioned without the user noticing what is happening at the IT infrastructure level, such as when servers are started or stopped. The application is just always there.
The migration has improved performance and security while reducing hosting costs.