Unit:
Κατεύθυνση Υπολογιστικά Συστήματα: Λογισμικό και ΥλικόΠληροφορική
Author:
Giannakouris-Salalidis Viktor
Supervisors info:
Ιωάννης Ιωαννίδης, Καθηγητής, Τμήμα Πληροφορικής και Τηλεπικοινωνιών, Εθνικό και Καποδιστριακό Πανεπιστήμιο Αθηνών
Original Title:
A Sample Index for Approximate Query Processing
Translated title:
A Sample Index for Approximate Query Processing
Summary:
We introduce Sample Index, a novel index structure that aims to enhance the sampling performance in a database system. Our idea is based on the observation that the overheads resulting from the expensive sampling steps during query execution can be mitigated by leveraging an index that is created offline. Our index is able to serve \emph{fresh} samples even when queries are issued during continuous inserts in the database, i.e., during an ETL process. Our experimental evaluation proves that our sample index implementation in MonetDB can achieve performance improvements ranging from 2x to 4.5x better query execution times.
Main subject category:
Technology - Computer science
Keywords:
Database System,Sampling,Index