A Comparative Study of Log-Only and In-Place Update Based Temporal Object Database Systems

Abstract

In most current database systems, data is updated in-place. In order to support recovery and increase performance, write-ahead logging is used. This logging defers the in-place updates, ho w ever sooner or later, the updates have to be applied to the database. This often results in non-sequential writing of lots of pages, creating a write bottleneck. To avoid this, another approach is to eliminate the database completely, and use a log-only approach. The log is written contiguously to the disk, in a no-overwrite way, in large blocks. The log-only approach is particularly interesting for transaction-time object database systems (TODBs). While previous approaches to TODBs have been page based, i.e., when an object has been modified, the whole page the object resides on has to be written back, our approach is object based. One of the objections against operating at object ragularity is that the read cost will be prohibitiely high. We will in this paper show that this is not necessarily true. We use analytical cost models to compare the performance of log-only and in-place update TODBs, and the analysis shows that with the workload we expect to be typical for future TODBs, the log-only approach is highly competitive with the traditional in-place update approach.