The big challenge of the solution was to create and maintain a distributed, scalable, independent and redundant environment.

The concept of a distributed and independent environment is to have a group of elements responsible for each geographic region, thus mitigating any unavailability and also making it faster to query data by region.

Two main points of user management and access plans that are redundant to each other and that replicate information in realtime to all secondary nodes (in each region) automatically.

Management of users and access plans was initially done through a layer of web services created especially for this. Later the direct access to the sql database allowed more agility by the legacy system.

Through several accounting servers, one in each microregion, querying a user's accounting can be a problem if you don't know which element they authenticated to and started accounting.

To resolve this issue, I implemented a data consolidator server and applications that collected in real-time any changes in user sessions, including storing open sessions.

All network elements allowed connection from any router, even remote, via the internet. In this way, each router could point to 2 or more authenticators and in case of failure of one, the other was automatically set as default/main, avoiding unavailability in new accesses.

In order to reduce the database overhead due to the intense I/O of realtime replications and accounting consolidations, I created resources to automatically summarize user traffic by day, month and year. So, to know, for example, the top 10 users in downloads in a given month, the information was already available in the database, as the data was summarized and consolidated in realtime in the main database.

Given basic and advanced training on the Radius environment and functionality to the NOC and Network Engineering team.

View here the basic concept of radius developed.