Cdn A Resource Controller, The Design And Realization

The requirement of video has become urgent than ever recently. However, the video data takes up too much internet resource due to its big volume. To provide faster access to users and to lower the rate of network congestion, bringing in a middle layer CDN into the current Internet is a most popular strategy to optimize the network. It ensures users can get what they need from a place that's nearest to them.The usual CDN network adopts the C/S-based content delivery mechanism. In this way, the source server takes much of the contents distributing loads, and the loads become larger with the number of edge servers that participate in the distribution increases. Distributing large-volumed and many video files tend to make the servers unbearable easily.To address the above problem, this thesis provides a mechanism that adopts contents distribution by P2P within the CDN network. In the proposed stratigy provided in this thesis, the way of dividing the contents distributing loads into all the nodes(including central and edge nodes) in the CDN network is adopted. Distributing in this way, the source and edge servers all become peers. The edge server is not only the client of contents distribution, but also the server that distributes, so through the implementation of this thesis, the overall load-balancing ability of the CDN network is improved to some point.Then the focus is moved to the design and implementation of a core scheduling part of the system-the resources controller. It's the module of resource controlling and scheduling, is the central server in the proposed system and is in charge of the searching and querying of the various servers'info. It collects info from ESD and TS, provides service to CP, TS and CR. To achieve such functions, RC needs to interact with many other communicating entries. It's designed that one thread can interact with many other communicating entries and these interactions are going with non-blocking mode to improve the capability. The business logic is abstracted into the TASK class, which includes the state machine of the business. The workflow of the business is controlled by the function GOTONEXTSTATE. The states of TASK are grouped by the IO incidents, so they can be driven by EPOLL. The specific driving of the business is accomplished by the functions of READFROM and WRITETO. The workflow processing of one TASK is involved by many AGENTs, and the transitions among different AGENTs are controlled by the TASK's state machines. Then the system functions of the whole system are finally achieved by designing the interfaces and communication protocols between RC and other relevant modules.At last by designing and conducting specific tests, it's proved that the RC server has fully achieved the pre-designed functions and has taken the responsibilities of controlling and acheduling the resources.