Dynamic Compression Scheduling for Content Delivery Networks

Abstract

Content compression is one of the most important steps for fast delivery of content to the client. In the existing Content Delivery Network (CDN) system, content compression is performed on edge servers. With the increase in the number of customers using Medianova CDN, a significant increase was observed at CPU usage in the edge servers. This leads to the increase in the CDN response time to the user and decreases in the performance. When the problem is examined, it is observed that about 40% of the CPU's usage is caused by compression. CPU compression is reduced to approximately 15% when the compression is transferred to the Origin-Shield servers which is a higher layer in the CDN architecture instead of the edge servers. However, because content protection servers serve more than one edge server, it may also cause the edge servers to fail to service when the server is overloaded. Therefore, compression must be performed dynamically on both edge and Origin-Shield servers. In this study, a method that dynamically manages the servers in which compression will take place is proposed. In this way, bandwidth between Origin-Shield and edge server will be used more efficiently and CDN performance will be improved.