Hybrid Load Scheduling in Content Delivery Network Comprising of Under Populated Clusters

Serving content to the worldwide distributed clients with high availability, high efficiency and minimum delay is the primary challenge for today's virtual world. Content delivery network (CDN) was evolved with the aim to send the data at the client's doorstep with almost zero latency. CDN disseminates the content at various edge servers deployed to the client's close proximity. But locating the positions for deployment and also sharing the load among the servers are two major concerns for CDN. In this paper, unsupervised K-means clustering is considered for selecting the locations for server deployment which also considers the under populated clusters based on a parameter called population threshold. This paper introduces a hybrid load sharing model which is mainly concerned about the traffic coming from the clusters with population lesser than the threshold. The result of the study shows that this hybrid approach enhances the server utilization factors of the surrogates deployed in the network while minimizes the server maintenance and cost overhead.