Dominik Durner, Lennart Espe, Jana Giceva, Anja Gruenheid
CIDR 2024, Conference on Innovative Data Systems Research
With the shift to databases-as-a-service, vendors are able to collect high-level database traces of executed workloads while retaining the privacy of their customers. In contrast to pure end-to-end latency statistics, traces contain enriched information that is useful for tasks such as workload monitoring and regression testing. Despite its importance, efficient analysis and exploration of traces and their rich feature space remains a challenge. In this paper, we introduce TracEx, an open-source Trace Exploration tool that facilitates workload trace analysis and comparison for database systems. TracEx allows users to understand their workload by providing an intuitive, visual interface that explores the workload along different dimensions, e.g., resource utilization or database operator usage. Additionally, users are able to contrast and compare workloads that have been collected from different hardware configurations or even compare traces between database systems.