The massive scale of Cloud computing has enabled the popularity of the internet and growth of multimedia streaming. Live streaming applications have greatly benefitted on deploying it in the cloud environment due to its abundant availability of resources and features that handle scalability with agility. Managing Device heterogeneity is critical and affects user experience drastically. Response time and bandwidth are other issues to be focused on. The streaming services involve both desktop users and mobile users. High-definition video applications are often challenging for mobile devices due to their limited processing capability and bandwidth-constrained network connection. Cloud computing environments are elastic in nature which balances the load according to the fluctuations in the network. It is also easy to re-commission the required services in case of failures. But downgrade of services is possible. The reasons for the downgrade of services are many however hardware failure is one of the chief causes and queue overflow during re-commissioning is another. Whatever the cause, the effect of downgrade is drastic on the customer. To meet up with the problem of resource allocation, bandwidth allocation and fault tolerance issues and at the same time to guarantee the desired level of Quality of Experience (QoE) to the end-users, an entire framework is proposed with novel algorithms for all the addressed issues. The resource allocation, performed at the cloud end needs to be dynamic. Our framework incorporates the proposed Guess Fit algorithm to provision virtual machines dynamically based on priority scores calculated using probabilities as a result of the combined Naïve Bayes algorithm with association rule mining. The scores also take into account the hit ratios and the penalty values. It is found to perform better than the existing Fit algorithms. On the client end, an efficient cluster bandwidth allocation algorithm (CBA algorithm) is proposed to share bandwidth resources among the fixed device and the mobile devices in a cluster. The framework also incorporates a switching table-based fault tolerance module. The switching table is entirely built based on AND-OR logic and help in desktop migration for uninterrupted streaming.