Heuristic algorithm for periodic patterns discovery in a database workload reconstruction

Prelegent(ci)

Marta Jadwiga Burzańska

Afiliacja

UMK

Język referatu

polski

Termin

29 kwietnia 2025 10:15

Link

https://meet.google.com/wtn-vmev-myx

Informacje na temat wydarzenia

Online talk

Tytuł w języku angielskim

Heuristic algorithm for periodic patterns discovery in a database workload reconstruction

Seminarium

Seminarium "DeSeR: Dane, strumienie, rozpraszanie"

---

Information about the existence of periodic patterns in a database workload can play a big part in the process of database tuning. However, full analysis of audit trails can be
cumbersome and time-consuming. This talk discusses a heuristic algorithm that focuses on workload reconstruction based on pattern discovery in a simplified workload notation. This notation is based on multisets representing database actions (such as user queries) requiring access to specific persistent objects, but without the access cost analysis. Each action in this notation is a multiset of accessed objects, which can be tables, system files, views, etc. The theoretical model for such an approach has been discussed in detail in the authors’ previous work. This work is mostly proof-of-a-concept for the theoretical approach. Additionally, in order to test the performance of the proposed algorithm, a testdata generator has been constructed. Both the previous and the current results are parts of a research project dealing with the application of periodic pattern theory to the field of database optimization and tuning

---