Internet plays important role in human lives. It becomes integrated, inseparable part of our daily routine we cannot imagine a day without using Internet. Internet is used for many purposes fun, shopping, communication, research, study. The technology has emerged so rapidly since it became public in early 1990s. According to the Internet World Stats 17.8% of world population is using Internet; another interesting aspect is that Usage growth is 225% in the world in general. Day by day web servers have to support more and more clients, but most of the systems cannot keep up with this growing demand. This continuous growth in the number of Internet users often results in popular web sites becoming overloaded. This might happen, for example, when a web site is linked by some very popular web site such as Slashdot, or when web site's link is advertised to a wide public in the media. In both cases, the number of requests received by the web site grows rapidly, causing the server's capacity to become overloaded. Overloaded web sites proceed as many requests as possible, and simply drop the remaining ones. Such events are often referred to as Slashdot effects, hot spots, or flash crowds.
Main topic of our research is the flash crowds and Internet systems related to alleviation of the flash crowds. The flash crowd is a rapid increase in traffic load to some particular website. This rapid increase causes the website to become unreachable, leaving the clients with unsatisfied requests. Flash crowds present a significant problem to web site owners. In the case of commercial web sites, a flash crowd can lead to severe financial losses, as clients often resign from purchasing the goods and switch to another, more accessible web site. However, non-commercial web sites can also experience flash crowds. In general, it is impossible to predict which sites shall be subject to flash crowds, and so each web site should be ready to handle them. The mostly used method in alleviation of the flash crowds is distributed infrastructure of surrogate servers called Content Distribution Network (CDN).
The CDN is a widely used Internet technique to improve the performance and availability of web sites by deploying geographically dispersed surrogate servers and distributing client requests to an "appropriate" server based on various considerations. This effectively increases the client-serving capacity of the web site by that of the CDN, which enables the web site to service all the clients with a good performance.
The main shortcoming of related researches is that they do not support dynamic resizing feature of a cloud of the surrogates; all surrogates in the cloud are involved in the alleviation process from the beginning. Our system, Flash Crowds Alleviation Network (FCAN) is a system to provide resources to web sites to overcome flash crowds. A main feature of FCAN is its dynamically resizing feature, which can adapt to request load of flash crowds by enlarging or shrinking a cloud of surrogate servers used by the web sites. Moreover it has the multi-server support feature, with the help of this feature the FCAN system can be used by several servers experiencing flash crowds events simultaneously. To avoid situation when resources of the system are used by several servers, we implement priority tables. Every resource in FCAN is assigned priority values, which are different for every server, it is done to avoid collisions; so if possible resource is used in only one alleviation procedure.
Simulation results showed that system with new modifications performs very well. Dynamic resizing feature provide more flexibility for the system and servers in need. Flash crowd alleviation procedure may start with small subset of proxies involved, and then it may grow or shrink according to the load. Multi-server support feature raise system's limits; several servers can benefit from the system at the same time. With the help of priority table we can prevent collision, by avoiding proxies to be used in several flash crowds' events.