A large U.S. telecommunication company operates a system for detecting fraudulent calls from cell phones. The system is required to monitor call events and watch for unusual activity such as cell phone cloning (duplicating another cell phone’s connection signature).
In Figure 1, call detail records (CDRs) are sent from fraud databases in a subordinate call center hosted on numerous HPE NonStop servers, to the master fraud database that hosts a fraud application and maintains a repository of all cell phone IDs. These IDs can be compared against incoming and outgoing call signatures. As calls are made, the call-processing nodes send the cell phone signatures for comparison to the IDs held in the master fraud database, which checks for fraud, cell phone cloning, etc. It then tells the cell-processing nodes to suspend suspicious cell numbers.
- The first step is to create a database of cell phone accounts and call events in a master fraud database, which is an intensive transaction-processing application.
- Complex fraud queries are run in real-time against the call events in this database and those results are returned to the master fraud application.
- To offload query processing from the system that was recording call events, the company uses an asymmetric capacity expansion configuration.
- A master NonStop node records all of the call events, runs the fraud application, and replicates suspicious cell phone numbers to the processing nodes in the call center.
- These nodes compare fraud detection queries to the call events, in order to detect unusual call activity.
- If suspected fraudulent activity is detected, the master fraud system is notified so that it can take appropriate action.
- The query nodes support a customer call center that can handle customer calls concerning actions because of suspected fraudulent activity.