Research On The Key Technologies Of Content Delivery Network And Its System Design

Content Delivery Network (CDN) is an accelerating service for websites, which reducesthe bandwidth cost with cheap caching storage and accelerates speed of access by deployingdistributed servers across the Internet. However, with the expansion of website sacle andincreasing number of users, the server placement of new CDN node becomes complicated anddifficult. And the current GSLB strategy based on DNS redirection is inaccurate, whichaffects the speed the of CDN.This paper primarily focues on the two issues above and has achieved an open sourceCDN system, a method of server placement for CDN node based on Geo-IP clustering and aGSLB strategy based on active speed detection. The major work of this paper is as follows.1. Design and implementation of the open source CDN system: given that the currentCDN costs dearly, this paper first explores the basic functions and criticaltechnologies and then works out an open source CDN system with three layers, thusreducing the construction cost of CDN. For the big data produced by CDN servers, wehave designed a distributed processing system based on Hadoop, which minimizes thetime of processing big data with correct results;2. Research of key technologies: the innovation point of this paper lies not only in theimplementation of CDN, but also in the optimization of the two major issues, i.e. theserver placement for new CDN node and the GSLB strategy. The key technologies arelisted below.(a) Geo-IP clustering: in face of the complexity of server placement for CDN newnode, we process the logs of CDN on Hadoop system and have achieved a methodof server placement based on the clustering of users’ real IP, which reflects theclustering center of users.(b) Active speed detection: due to the inaccuracy of GSLB strategy based on DNSredirection, we have achieved a load balancing strategy based on the active speeddetection, which utilizes the history access of users and increases the accuracy ofGSLB. 3. Test of CDN and its key technologies: the results show that our open source CDNoutweighs the current ones in costs and features; the method of server placementbased on Geo-IP clustering has a wider applicability and higher accuracy; the GSLBstrategy based on active speed detection has a stronger adaptability and higheraccuracy when the IP location of user is unknown or the user have visted before.