Hospital database workload and fault forecasting

With the growing importance of hospital information systems, databases became indispensable tools for day-to-day tasks in healthcare units. They store important and confidential information about patients clinical status and about the other hospital services. Thus, they must be permanently available, reliable and at high performance. In many healthcare units, fault tolerant systems are used. They ensure the availability, reliability and disaster recovery of data. However, these mechanisms do not allow the prediction or prevention of faults. In this context, it emerges the necessity of developing a fault forecasting system. The objectives of this paper are monitoring database performance to verify the normal workload for the main database of Centro Hospitalar do Porto and adapt a forecasting model used in medicine into the database context. Based on percentiles it was created a scale to represent the severity of situations. It was observe that the critical workload period is the period between 10:00 am and 12:00 am. Moreover, abnormal situations were detected and it was possible to send alerts and to request assistance.