Target-based database synchronization

Abstract

Synchronizing source and target databases is an important task in many database applications. There are instances when the synchronization of source and target databases must be driven from the target's side and involve no changes to the source's schema or triggers. We describe an algorithm for such synchronization. Our algorithm groups tuples into partitions and compares hashes of matching source and target partitions before synchronizing only those partitions whose hashes do not match. The hash comparisons decrease the number of tuples that must be exchanged when the source and target are nearly synchronized already. Two variants of full replication that differ on locking strategies are used as benchmarks. Empirical results show that our method outperforms both when there are few changes to the database and outperforms row-level locking when fewer than 70% of the partitions are changed.

Data Synchronization Model for Heterogeneous Mobile Databases and Server-side Database

“…The target database can be either client or master database [9]. Whereas in others such as [2], [14] and [15], the process is initiated by the owner of changes, i.e.…”

“…Amongst them are [6], [7] who proposed Synch Algorithm using Message Digest (SAMD) and [8] who introduced a stateful DS, all for the purpose of minimizing the load on the mobile devices. Also, [9] suggested a target based algorithm which always initiate the synchronization process from the target database.…”

Target-Based, Privacy Preserving, and Incremental Association Rule Mining

“…This transition features change that occurs across source sites. An efficient way to track changes in very large volumes of transactional data is proposed in [3]. Known as Target-Based Database Synchronization (TBDS), this approach requires a target system to synchronize itself with the source systems by tracking only the differential in large source systems.…”

“…For an input represented by a list of 2 n numbers, the Haar wavelet transform simply pairs up input values, storing the difference and passing the sum. The difference is called the detail coefficients and the sum is called the approximation coefficients 3 . This process can be repeated recursively over approximation coefficients in the previous level.…”

“…DWT allows a multi-resolution representation of discrete data. Important patterns in the high resolution data are usually preserved in the low 3 We divide the sum and difference by two in this paper.…”