Pruning Update Log Files in Intermittently Connected Databases

Modern databases allow mobile clients that subscribe to replicated data, to process the replica without requiring continuous connectivity, and to receive the updates while connectedto the server. In such an environment--usually known as the Intermittently Connected Database (ICDB) system--the server should maintain the updates to the database in the log file(s). These update log files should be pruned to reduce update retrieval time. In this paper we propose two pruning algorithms, based on the periodic connectivity of the clients, that consider two scenarios: uniform client connectivity patterns and widely varying client connectivity patterns. In the former case, the complete pruning algorithm is effective in keeping the log file size within a bound, hence reducing both disk I/O during update propagation, and disk storage space; whereas, in case of the latter, the partial pruning algorithm achieves significant further reduction in disk I/O while retrieving the updates. Any reduction i...

• Type: Conference
• Year: 2004
• Where: WIS
• Authors: Liton Chakraborty, Ajit Singh, Kshirasagar Naik
• Keywords: Client Connectivity Patterns, Disk I/o, Information Technology, Pruning Algorithm, WIS 2004