• Cezar Chitac
  • Robertas Kerpys
  • Raluca Marcuta
1. semester, Datalogi, Kandidat (Kandidatuddannelse)
There is a growing need to accurately track moving objects in a given area. In order to support efficient queries on positions of tracked objects, a customized spatial data structure is used. The application domain requires efficient updates to be performed in order to maintain the reported position of objects as fresh as possible. With the evolution of main memory databases this goal becomes easier to achieve. However, in order to take full advantage of highly parallel modern CPUs, one must take into consideration the issue of concurrency in such a system. Being a fact that locking and, or latching greatly affect the behavior of the system with regards to concurrency between operations, the goal of the present paper is to introduce a concurrency algorithm that avoids locking and latching as much as possible. Based on this, two new approaches are presented that ensure high concurrency when structural modifications due to underfull and overfull nodes are taken into account.
Udgivelsesdato10 jan. 2011
Antal sider19
ID: 43732815