Load balancing approaches for user data writing and reading in cloud environment and a new approach

Cloud computing has become a magnificent working area both in academical and industrial researches. The solved and unsolved problems in Distributed Systems are moved into the Cloud Computing area. Distributing the load on to the whole system in a balance, levarages the performance that is called load balancing in these kind of systems. Load balancing is still a niche working area for Cloud Computing data centers. Load balancing approaches are reviewed and especially the most promising non-transactional methods are tested in a cloud environment. Also in the study, a successful sharding column selection for different data types and the effect of this selection on load balancing and the corresponding performance is examined. The Screen Tracker application, developed to increase the productivity of the employees, is used for aforementioned tests and examinations. This application is developed by Sword-IT and aims to increase employee productivity by collecting and analyzing the data about how much time the employees spend at work by using their computers. The application is designed by taking advantage of load balancing approaches during the writing and reading of application-generated data for the purpose of storing employee data offline and then querying online. A distributed structure will be created to record the data from the screen tracker application. In this structure, one coordinator node, new load nodes that can be clustered according to the load case, retention nodes to hold the data, and read end-point nodes to read the data will be constructed. Detailed information about the operation of the Screen Tracker application and its architecture is included in Chapter 4. PostgreSQL Database Management System is used as a Database Management System within the Project. The application data is sharded by a column and distributed more than one server in terms of scaling and load balancing and then performance affects of different topologies are compared. Used methodologies and approaches in the industry and academic researches are reviewed and the most promising ones are tested and compared. Also the performance gain according to round robin is evaluated during the work.