TY - GEN

T1 - The study of multidimensional R-Tree -based index scalability in multicore environment

AU - Smirnov, Kirill

AU - Chernishev, George

AU - Fedotovsky, Pavel

AU - Erokhin, George

AU - Cherednik, Kirill

PY - 2014

Y1 - 2014

N2 - In this paper we consider the scalability issues of a classical data structure used for multidimensional indexing: the R-Tree. This data structure allows for an efficient retrieval of records in low-dimensional spaces and is de facto standard of the industry. Following the design guidelines of the GiST model we have implemented a prototype which supports concurrent (parallel) access and provides read committed isolation level. Using our prototype we study the impact of threads and cores on the performance of the system. In order to do this, we evaluate it in several scenarios which may occur during the course of DBMS operation.

AB - In this paper we consider the scalability issues of a classical data structure used for multidimensional indexing: the R-Tree. This data structure allows for an efficient retrieval of records in low-dimensional spaces and is de facto standard of the industry. Following the design guidelines of the GiST model we have implemented a prototype which supports concurrent (parallel) access and provides read committed isolation level. Using our prototype we study the impact of threads and cores on the performance of the system. In order to do this, we evaluate it in several scenarios which may occur during the course of DBMS operation.

KW - threads

KW - scalability

KW - databases

KW - multidimensional indexing

KW - in-memory index

KW - R-Tree

KW - GiST

KW - experimental evaluation

M3 - Conference contribution

SP - 192

EP - 203

BT - 9th Ershov Informatics Conference Preliminary

PB - Прайс-курьер

ER -