الفهرس 
Chapter 1: INTRODUCTIONOnly 14 pages are availabe for public view
 |  | from | 102 |  |  |
| | | from | 102 | | |
Abstract
Concurrency has its importance from its need in many areas such as banking
systems, websites developing, stock exchange systems, and other applications,
which based on working with database system. Concurrency has a lot of
advantages such as improving response time, parallel execution of transactions,
improvement of system speed and enhancement of database system. However,
there are some disadvantages such as affecting on database consistency and data
integrity. There are some problems of concurrency such as deadlock, lost updates,
transactions conflict and dirty read. These problems effects on database and
transaction performance. To solve these problems, concurrency should be applied
under control. So, it is necessary to enhance concurrency control algorithms to
solve these problems.
This thesis presents three proposed algorithms, The first proposed algorithm is
“Cell lock”. This algorithm presents an efficient method for enhancing the number
of concurrent committed transactions, throughput of database system and reducing
the number of rolledback ones for concurrency control algorithms. This proposed
algorithm is based on applying data blocking. It works by applying locking on restricted point of database called “Cell”. The number of committed transactions is
enhanced by reducing data blocking area and time. It is tested and compared with
other locking levels of two phase locking algorithm. Experimental results
demonstrate that the “cell lock” algorithm is very efficient in enhancing the
number of committed transaction, Throughput is increased and the number of
rolledback transactions is reduced. But, it effects the data consistency and
accuracy, so it is improved via the second proposal “Modified Cell Lock”. This
proposal aims to ensure data consistency while keeping high number of committed
transactions on database. Experimental results demonstrate that the second
proposed algorithm is very efficient in enhancing the number of committed
transaction. It also, enhance the throughput and reduces the number of rolledback
ones, but it suffers from deadlock problem.
The last proposed algorithm is “Deadlock Free Cell lock”. This algorithm
contributes also for enhancing database and transaction performance by applying a
method to solve the deadlock problem of the “Modified Cell Lock” algorithm.
This thesis presents a concurrency control algorithm -Deadlock Free Cell Lockwith a high number of committed transactions, a low number of rolledback
transactions, deadlock free, and high data consistency. Finally, a comparative study
is employed to compare the three proposed algorithms with other traditional
concurrency control algorithms proved the effectiveness of the proposed
algorithms. DFCL improves the average number of committed transactions by
62.9% and reduces the average number of rolledback transactions by 9.6% and the
average number of waiting transactions by 53.3% compared with Two Phase
Locking Algorithm.