A Virtual Machine Based Dependability Validation Mechanism For Cloud Computing Systems

Cloud computing has become one of the major technology trends in the field ofinformation technology. Cloud computing systems are built on top of virtualizationtechnology, and are dynamically expanding all the time. As a result, the services deployedon them have remarkable differences with the traditional desktop software or the servicesdeployed in distributed systems. Software testing is playing an increasingly important rolewhile the LOC (Lines of Code) and complexity of software increasing. Some methods orproducts have been proposed to improve the efficiency of testing. Their basic flow isdefined as: build a test bed and then transmit test commands to targeted nodes to evaluatethe dependability of the system, using the test bed. These solutions give little considerationon the efficiency and the changes of environment because they are proposed for stand-aloneor distributed systems. The facts above also make these solutions neither able to test andanalyze thousands of nodes simultaneously, nor able to handle the dynamically changingnetwork and hardware facilities in cloud computing systems. What’s more, these solutionsare pre-loaded with some test modules and are not able to mount any test modules at anytime, leaving them not able to test and analyze the dependability of cloud computingsystems comprehensively. It has been regarded as a great challenge to test the dependabilityof cloud computing systems with a high accurate, a good extensibility and efficiency whiletargeting tested nodes.In this work, we present the design, implementation, and evaluation of CloudDTB, atest bed for testing the dependability of cloud computing systems. First, CloudTB is basedon a highly scalable and efficient communication mechanism and is able to guarantee aminimal time spent on sending test commands to targeted nodes. Second, with the help of aSSH (Secure Shell) and Hash based verification mechanism, CloudDTB is able toaccurately locate a targeted nodes in the dynamically changing cloud computing systems.Third, CloudDTB provides a universal strategy to mount any test modules and edit as manytest cases as needed to provide a good expandability. In the end, CloudDTB provides amulti-layer log library to give a better description on test information.The evaluation results show that CloudDTB is able to send100thousand and receive 40thousand pieces of1KB-length message per second while testing millions of nodessimultaneously, with an accuracy of100%for targeting and a good scalability. Besides,CloudDTB is of good expandability to mount different test modules and test those usingdifferent test cases. Finally, CloudDTB is able to generate detailed descriptive informationabout the testing.