The cloud computing paradigm enables elastic resources to be scaled at run time satisfy customers' demand. Cloud computing provisions on-demand service to users based on a pay-as-you-go manner. This novel paradigm enables cloud users or tenant users to afford computational resources in the form of virtual machines as utilities, just like electricity, instead of paying for and building computing infrastructures by their own. Performance usually specified through service level agreement performance commitment of clouds is one of key research challenges and draws great research interests. Thus, performance issues of cloud infrastructures have been receiving considerable interest by both researchers and practitioners as a prominent activity for improving cloud quality. This work develops an analytical approach to dynamic performance modeling and trend prediction of fault-prone Infrastructure-as-a-Service clouds. The proposed analytical approach is based on a time-series and stochastic-process-based model. It is capable of predicting the expected system responsiveness and request rejection rate under variable load intensities, fault frequencies, multiplexing abilities, and instantiation processing times. A comparative study between theoretical and measured performance results through a real-world campus cloud is carried out to prove the correctness and accuracy of the proposed prediction approach.