Concurrency control performance modeling: alternatives and implications
ACM Transactions on Database Systems (TODS)
801 storage: architecture and programming
ACM Transactions on Computer Systems (TOCS)
Computer
Performance Considerations for an Operating System Transaction Manager
IEEE Transactions on Software Engineering
Architecture support for single address space operating systems
ASPLOS V Proceedings of the fifth international conference on Architectural support for programming languages and operating systems
Lightweight recoverable virtual memory
SOSP '93 Proceedings of the fourteenth ACM symposium on Operating systems principles
Tempest and typhoon: user-level shared memory
ISCA '94 Proceedings of the 21st annual international symposium on Computer architecture
Hardware and software support for efficient exception handling
ASPLOS VI Proceedings of the sixth international conference on Architectural support for programming languages and operating systems
SOSP '95 Proceedings of the fifteenth ACM symposium on Operating systems principles
Improving applications performance: a memory model and cache architecture
ACM SIGARCH Computer Architecture News
IEEE Micro
Virtual memory transaction management
ACM SIGOPS Operating Systems Review
Multiview model for protection and access control
Multiview model for protection and access control
Multi-view memory to support OS locking for transaction systems
IDEAS'97 Proceedings of the 1997 international conference on International database engineering and applications symposium
| Hi-index | 0.00 |
The focus of this paper is to investigate the use of the Multi-view Virtual Memory (MVM) in providing efficient locking services for transaction processing or DB systems. The granularity of locking is customizable by applications and, furthermore, the locking granularities for read and write operations may differ.The system provides for enforcement of access control protocols on units of a memory region through FSM specification. The units of data can vary in size from one region of memory to another. Threads executing transactions do not explicitly request locks on data items - they simply access the data items while locking is represented through state transitions and is performed automatically. Only when a thread is suspended are the state changes communicated to the software lock manager. Context switching overhead is thus reduced.Synchronization of threads is possible within a single task or across tasks. Any synchronization protocol that can be represented by an FSM is supported by the MVM system. This paper describes how MVM supports sophisticated locking in which the locking granularities for read and write operations differ. Benefits, which are expressed in terms of throughput, gained using such locking are investigated using a simulation study.