Transaction Time Support Inside a Database strongEnginestrong.pdfVIP

Transaction Time Support Inside a Database Engine David Lomet, Roger Barga Mohamed F. Mokbel ∗ German Shegalov ∗ Microsoft Research University of Minnesota Max Planck Institute Redmond, WA Minneapolis, MN Saarbrucken, Germany {lomet,barga}@ mokbel@ shegalov@mpi-sb.mpg.de Rui Wang ∗ Yunyue Zhu ∗ Northeastern University New York University Boston, MA New York, NY bradrui@ yunyue@ Abstract mercial database systems that support temporal functional- ity with historical versions organized for rapid retrieval, nor Transaction time databases retain and provide access with timestamping that is highly precise. Further, layering to prior states of a database. An update “inserts” a new temporal support on top of a database system is cumber- record while preserving the old version. Immortal DB some and typically is not practical [16, 35]. builds transaction time database support into a database In this paper, we present Immortal DB, our research engine, not in middleware. It supports as of queries re- prototype that provides transaction time database support turning records current at the specified time. It also sup- built into the SQL Server database engine, not layered on ports snapshot isolation concurre