Semantics-based optimization under Epsilon serializability
Wenwey Hseush · 1995
The recent movement to advanced applications and the continuous drop in hardware prices impact the design of on-line transaction processing (OLTP). The data contention problems imposed by serializability become an increasingly serious performance bottleneck. Many efforts have been made to decrease data contention in OLTP systems. In general, two directions of work have been studied. The first direction (semantics-based optimization) is to design practical optimization techniques while preserving serializability as the correctness criterion. The second direction is to relax the restrictions on serializability. A good example is the design of epsilon transactions that operate under the semantics of epsilon serializability (ESR). The goal of this thesis is to study the relationship between ESR and semantics-based optimization and design new methods to further improve performance. We show that ESR and several well...