Development of a server monitoring tool and defining performance thresholds

Sammanfattning: Today we depend on data based services, for example in logistics, at warehouses, and accounting services. When a data based service is delayed it can depend on its servers being overloaded, leading to economic losses. To prevent this, it is necessary to have appropriate monitoring tools that can make the system administrators aware of when performance thresholds have been breached and servers are reaching a critical state. The focus of this thesis was twofold: to define performance thresholds and to simplify monitoring of critical servers with presentation on a dashboard. As a result of this, two thresholds were produced. A threshold for high CPU utilization which further improved the existing performance thresholds by making the required time for triggering dependent on the current CPU utilization. This proved to give us a better result in 75% of the tested cases compared to the previous model. The second produced threshold was for low CPU utilization. Through studying servers with more allocated resources than they require it was possible to inform the system administrator to disable or reallocate those resources to another server with higher demand for resources. This reduces server costs and improves performance in some cases. The primary intention behind implementing a dashboard was to have a way of presenting when the defined thresholds had been breached. After evaluation of the dashboard through a usability test we came to the conclusion that all end-users were satisfied with the prototype and four out of five considered the dashboard to be a tool which they would use in the near future.