Sk911 M. Stonebraker and G. Kemnitz. the Postgres Next-generation Database Management System. Mn944 C. Mohan and I. Narang. Ariesscsa: a Method for Database Recovery in Client-server Architec- Tures. Local Disk Caching for Client-server Database Systems

A transaction recovery method supporting ne-granularity l o c king and partial rollbacks using write-ahead logging. MN911 C. Mohan and I. Narang. Recovery and coherency-control protocols for fast intersystem page transfer and ne-granularity l o c king in a shared disks transaction environment. Bil92aa A. Biliris. An eecient database storage structure for large dynamic objects. Bil92bb A. Biliris. The performance of three database storage structures for managing large objects. A study of three alternative w orkstation-server architectures for object-oriented database systems. 34 transactions, one of which is going to update some objects residing on the page. The recovery protocol was designed with the goal of minimizing the overhead of the recovery-related activities during normal transaction execution and the goal of ooering fast transaction abort and system restart times. In addition, we h a ve presented several performance studies of the EOS implementation of the con-currency control and recovery protocols that we h a ve described in the paper. From the results collected from these studies and from the limited number of published performance results for logging and recovery systems, we concluded that the overhead for many cases was reasonable, despite the write-intensive nature of the tests we ran. In addition, the performance study of the 2V-2PL concurrency control method showed that the concurrency level of the system increases considerably compared with the performance of the strict two-phase locking. Finally, the studies raised several issues that have to be addressed in order to improve the performance of the system, including: substituting the expensive UNIX semaphores with a faster test-and-set utility, eliminating the overhead during the write phase by batching reads from the transaction private log les so that the number of random seeks is reduced, and batching log writes to transaction private log les to avoid IIO overhead due to random seeks. EOS is used as the storage engine of the ODE OODBMS BGL + 93. The EOS facilities are also being used in a major AT&T project that provides interactive TV capabilities. This project requires eecient manipulation of multi-media objects in a client-server environment. EOS runs under UNIX on SPARCstation, Solaris, IBM RSS6000, and SGI architectures. EOS is written in C++ but it can also be used by programs written in C. Release 2.0 BP933 is available free of charge to Universities. We are currently working on issues related to inter-transaction caching, providing support for multiple servers and distributed transaction …