In recent years, there have been rapid growth in wireless technologies and by now we can provide much higher bandwidths and improved quality of service. In addition, a diverse selection of portable devices and services provided for them also makes the wireless networks more popular. However, we are not yet able to serve the increased traffic that comes with the heavier usage, and congestion can be detected in radio networks (especially in peak periods), and this may degrade the user experience. To treat this problem, new (enhanced) network management methods should be developed.
To address this problem we plan to establish a cache system which can reduce the network traffic (through caching data), and control the traffic so that it can be better adapted to the system load and available free capacities. To achieve this we would like to use traditional network management methods as well as apply a so-called pre-caching technique, where certain data objects would be loaded into the cache before they are actually consumed (this is called prefetching). This would be done in the less busy periods, and if we could predict future demands accurately, the bandwidth to transfer data for these can be saved for the future when network congestion might occur.
Firstly, I researched the related works, where I primarily focused on web and video prefetching. We concentrated on these areas, because it is reported that they are responsible for more than half of the traffic in wireless networks. In this process I got to know some system solutions that were proposed, and I found a number of algorithms that were developed to predict future user requests. Based on these information and our initial concepts I established the architecture and operational principles for the target system. The structure and functions of the system were implemented in a way that it can demonstrate the use cases that we devised for it.
In our system, we assumed that the clients use wireless network to access the Internet, and all their requests go through a proxy server. The radio connection represents the bottleneck in the network, and our task is to control the traffic flow efficiently. To do this, caches are placed at the clients and the proxy, and an additional central unit in the proxy will help regulating the pre-caching function in accordance with the network conditions, thereby adjusting the traffic usage. We managed to implement a prototype system that is able to control and schedule the radio traffic in accordance with the network conditions.