In this thesis, we present a simple and efficient recovery method, called ARIES/SR(ARIES/Selective Redo), which is based on ARIES(Algorithm for Recovery and Isolation Exploiting Semantics). ARIES performs redo for all updates done by either nonloser transaction or loser transaction, and thus significant overhead appears during restart after a system failure. To reduce this overhead, we propose ARIES/SR recovery algorithm. In this algorithm, to reduce the redo operations, redo is performed, using log record for updates done by only nonloser transaction.
ARIES/SR supports fine-granularity(e.g., record) locking and recovery using writeahead logging(WAL). All updates of a transaction are logged, including those performed during rollback. By appropriate chaining of the log records written during rollbacks to those written during forward process, a bounded amount of logging is ensured during rollback even in the face of repeated failures during restart or of nested rollbacks. ARIES/SR uses a log sequence number in each page to correlate the state of a page with to logged updates of that page.
Also in this thesis, we implement ARIES/SR based on WISS(Wisconsin storage system). We also present a basic y performance analysis of the implementation.
