Concurrency control in complex information systems: a semantics-based approach

CONCURRENCY CONTROL TECHNIQUES ADOPTED IN NEXT GENERATION INFORMATION BANKS SHOULD BE ABLE TO HANDLE HIGH LEVEL OPERATIONS ON ARBITRARY OBJECTS WITH COMPLEX STRUCTURE. UNFORTUNATELY, CURRENT APPROACHES TO CONCURRENCY CONTROL ARE HIGHLY INEFFICIENT FOR THESE TYPES OF OBJECTS. THIS DISSERTA- TION PROVIDES NEW TECHNIQUES FOR HIGH PERFORMANCE SEMANTICS-BASED CONCUR- RENCY CONTROL SCHEMES IN `COMPLEX INFORMATION SYSTEMS''. OUR TECHNIQUES MAKE USE OF THE AVAILABLE SEMANTIC INFORMATION IN SCHEDULING OPERATIONS. USING THE SEMANTICS OF THE OEPRATIONS, WE DEFINE A NEW NOTION OF CON- FLICT, WEAKER THAN COMMUTATIVITY, CALLED `RECOVERABILITY''. WE PRESENT A CONCURRENCY CONTROL SCHEME THAT USES RECOVERABILITY AND ALSO PRESENT A MULTILEVEL CONCURRENCY CONTROL PROTOCOL THAT USES RECOVERABILITY. IN ADDI- TION, AN ANALYSIS OF THE STRUCTURE OF THE HIGH LEVEL OPERATIONS IS USED DEFINE THE NOTION OF `RELATIVE CONFLICTS''; IT IS SHOWN THAT MULTILEVEL CONCURRENCY PROTOCOLS USING RELATIVE CONFLICT AS A BASIS FOR SCHEDULING OPERATIONS PRODUCE HIGH GAINS IN CONCURRENCY. WE VERIFY THE CORRECTNESS OF THE PROTOCOLS PRESENTED USING A MULTILEVEL GRAPH MODEL TO SHOW `SEMANTIC SERIALIZABILITY''. THE PERFORMANCE OF THE PROTOCOLS, BASED ON SEMANTIC SERIALIZABILITY, IS DETERMINED BY CONDUCTING EXTENSIVE SIMULATION STUDIES. BY TAKING ADVANTAGE OF SEMANTIC INFORMATION THAT IS AVAILABLE IN COMPLEX INFORMATION SYSTEMS, IN A GENERAL AND SYSTEMATIC WAY, OUR CONCURRENCY CONTR