A reliable and consistent fully distributed database
This paper presents an integrated concurrency and recovery algorithm. Strict timestamp ordering was used for concurrency control and incremental log with immediate updates for recovery. Strict timestamp ordering delays each write operation on a data item until there is no active transaction that iss...
Saved in:
| Main Author: | |
|---|---|
| Format: | text |
| Language: | English |
| Published: |
Animo Repository
1989
|
| Subjects: | |
| Online Access: | https://animorepository.dlsu.edu.ph/etd_masteral/1228 https://animorepository.dlsu.edu.ph/cgi/viewcontent.cgi?article=8066&context=etd_masteral |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | De La Salle University |
| Language: | English |
| Summary: | This paper presents an integrated concurrency and recovery algorithm. Strict timestamp ordering was used for concurrency control and incremental log with immediate updates for recovery. Strict timestamp ordering delays each write operation on a data item until there is no active transaction that issued a write on that file data. Reliability is achieved in the distributed database systems (DDBS) through replication and the recovery technique adopted. The algorithm allows a transaction to access any of the available copies of a data file. Since any copy of the data item can be accessed, data are available even in cases of site failures. The recovery technique ensures that each transaction would eventually terminate either by commiting or aborting. Moreover, it also takes care of restoring the database to a consistent state during recovery of a failed site. Consistency is achieved by controlling the accesses made to the database. This is solved by the concurrency control mechanism. Transaction restarts were minimized since strict timestamp ordering was used. The synchronization of the locals clocks was also included in the algorithm. The correctness of the algorithm was also proven and the message complexity of the algorithm was found to be of O (on), where a is a function of size of the queue used in the algorithm. A closed queueing network model was developed to evaluate the performance of the integrated algorithm. The performance measurements taken include response time, throughput and system utilization. The results indicate that the performance exhibited by the system is good even when subjected to heavy load. |
|---|