Nowadays, while large-sized multimedia objects are becoming very popular throughout the Internet, one of the important issues appears to be the acceleration of content delivery network (CDN) performance. CDN is a web system that delivers the web cached objects to the client and accelerates the web performance. Therefore the performance factor for any CDN is vital factor in determining the quality of services. The performance improvement can be achieved through load balancing technique, so the server load could be distributed to several clustered groups of machines and processed in parallel. Also the performance of CDN heavily depends on caching algorithm which is used to cache the web objects. This study investigates a method that improves the performance of delivering multimedia content through CDN while using RADS algorithm for caching large-sized objects separately from small-sized ones. We will also consider the efficient distribution of requests outgoing from local servers in order to balance the CDN load. This method uses various types of factors such as CPU processing time, I/O access time and Task Queue between nearby servers. At the end of the paper, we will see the experimental results derived from implementing the proposed optimization technique and observe how it could contribute to the effectiveness of CDN.