Using data similarity to achieve synchronization for free

Abstract

M.H. Graham (1993) proposed several conditions which are sufficient to guarantee that a transaction system will run serializably without any extra effort having to be taken. Systems satisfying these conditions are said to achieve serializability for free. The conditions considered by Graham are determined by a syntactic check on the transaction programs, and are independent of the semantics of data. In this paper, we use a semantic approach and propose a sufficient condition for achieving data synchronization for free, which is based on the concept of data similarity. Real-time transactions satisfying this condition can be scheduled correctly by any process scheduling discipline that is designed for the independent processes model, e.g. RMS and EDF algorithms, where no locking of data is assumed. The correctness of our approach is justified by exploiting the idea of /spl Delta/-serializability.<>