| Abstrak |
: |
Two new concurrency control algorithms are introduced for partially replicated distributed databases. They both maintain two values of a data item, and differ in that one requires all locks to be granted at one time, whereas the other does not. They are based on locking, and avoid deadlocks by using timestamps to establish an execution order when conflicts arise. Since they both proceed without any communication among schedulers, but only communication between the originating site and all participating sites, we say they are fully distributed |
