A heterogeneous multi-core architectural model for video scheduling for transcoding in clouds

A cluster of transcoding servers is essential for transcoding many on-demand videos. Cloud computing presents a scalable framework for online video transcoding, and the infrastructure as a service (IaaS) cloud provides heterogeneous virtual machines (VMs) for creating a dynamically scalable cluster of servers. Heterogeneous VMs consist of small or big cores, which are assigned dynamically to allocate varying sizes of videos to the appropriate VMs for transcoding. Earlier research has proposed cloud-based heterogeneous scheduling for allocating different types of videos to different types of VMs so that the quality of service is maintained by reducing video rejection. In this paper, we propose a heterogeneous multi-core video scheduling model that additionally estimates the number of VMs and cores per VM with the variation of the number of videos to optimize the resources and cost of a cloud-based transcoding system. We further estimate the model's overhead concerning the variation in the number of videos. We conducted experiments on random videos, and experimental results reveal that the proposed model provides an excellent estimation of the number of VMs and cores. The proposed model reduces the average cost by 5% and requires almost 10% fewer cores for processing video tasks than the existing work in average cases.