Self-healing framework for service level agreement monitoring and violation reacting in cloud computing

• Main Author: Mosallanejad, Ahmad
• Format: Thesis
• Language: English
• Published: 2014
• Online Access: http://psasir.upm.edu.my/id/eprint/60533/1/FSKTM%202014%2031IR.pdf; http://psasir.upm.edu.my/id/eprint/60533/
• Institution: Universiti Putra Malaysia

Service Level Agreement (SLA) is a mutual contract between service provider and consumer upon quality of service in cloud computing. A self-healing framework is unavoidable to monitor the agreed services and react against any probable SLA violation. Some SLA-based self-healing frameworks are presented but the rate of SLA violations is significantly high. Current SLA structure is not adapted for the hierarchical nature of SLAs in cloud computing. The response time of SLA monitoring systems also are not fast enough for early violation detection, and consuming a high overhead cost for bandwidth, CPU and memory consumption in both server and virtual machine sides. High reaction time for recovering violated services is also a considerate issue in this matter. Consequently, cloud consumers faced significant number of SLA violations in their services. The critical literature review conducted by this study highlighted the mentioned problem statements in detail. In this study, an extended SLA is proposed to fulfill the hierarchical structure of SLAs in cloud computing. The objective of self-monitoring SLA is mainly to reduce the monitoring response time and overhead. A self-healing framework is also proposed to reduce the reviving time for violated services. The proposed framework is developed based on self-monitoring SLA as to reduce the rate of SLA violations. The self-healing framework is evaluated by two different experiment scenarios in cloud computing. The proposed self-monitoring SLA and LoM2HiS, as a related work, are developed in the first experiment. Both of monitoring systems are executed to monitor virtual machines based on predefined SLA with the number of virtual machines increases from 1 to 4 units. The response time, bandwidth, CPU and memory consumption of monitoring systems are recorded and observed at run time. As for the second experiment, SLA1 and SLA2 are defined for hMailServer and SQL Database services respectively while SLA1 is depended on SLA2. The proposed selfhealing framework and three other alternative works are developed to keep hMailServer and SQL Server available. The implemented frameworks are executed in described scenario for 20 minutes and every 5 minutes an intrusion attack applied to stop SQL service. The reaction speed and the rate of SLA violation are measured in this experiment for results comparison. The response time of self-monitoring SLA is recorded to be seven times less than LoM2HiS. The average reviving time in self-healing framework is recorded to be two times lesser than developed related work based on LoM2HiS. The self-healing SLA also decreased the number of SLA violations. Therefore, the proposed self-healing framework is proven to reduce the overhead of SLA monitoring and the number of SLA violations in cloud computing. The proposed system has also been executed in real environment for validation purposes.