الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 114 |  |  |
| | | from | 114 | | |
Abstract
Cloud computing technology is a cost-efficient computing paradigm. Its paradigm delivers required computing resources for storing and managing information to customers through the Internet. In response to great success for cloud computing technology, many enterprises migrated their business-management software like Enterprise Resource Planning - ERP systems - and Enterprise infrastructure software like emails to cloud paradigm. Using applications or software as a service and the data explosion motivated deploying data management systems on cloud computing platforms. Building a relational cloud DBMS that combines high availability, scalability of NoSQL systems and the consistency and usability of traditional SQL databases, is the current challenge. Cloud DBMS needs linear scalability for serving large and continuously increased databases which can reach several terabytes.
This thesis proposes an architecture for cloud datacenters that provide database as a service. It achieves near linear scalability by separating processing pool from the storage pool and makes the latter shared between all database instances. In this architecture, adding a new database instance for a tenant does not require data migration between instances since the data is shared between all tenants. It also proposes a fragmentation algorithm for cloud shared storage DBMS to parallelize DBMS operations on all storage devices and achieves better performance especially for expensive transactions. Also, it exploits a storage management and monitoring module for better data placement that balance I/O load in ’read and write’ operations between these storage devices. It never allows skewness in storage utilization to happen. Putting related data on different storage devices allows parallel access to them. Finally this thesis proposed an optimistic concurrency control algorithm for the proposed shared storage DBMS.All these algorithms were implemented in C++. They have been evaluated with large real datasets. The results showed that the proposed algorithms enhanced the performance and scalability for the cloud DBMS.