Modeling and analysis of flash crowds in P2P-based live streaming systems

P2P-based live streaming has been widely adopted in live streaming applications, regarded as a promising technique. However, flash crowd is still one of challenges for P2P live streaming systems, which often occur when a great number of users join a channel simultaneously, typically during the first few minutes of a new released program. Under the flash crowd, the users are more likely to encounter a long startup delay and retry multiple times before they leave out of impatience. Currently, there are some analysis models on flash crowd, but none of them refer to the effects of user access rate control and bandwidth-ordered control on flash crowd. However, we guess the two aspects are vital to handle the problem. We develop two mathematical models to analyze the two aspects. Results of model analysis verify our guesses and find that both methods can alleviate the flash crowd. We also find the access rate must be controlled properly, neither too high nor too low. And we think all these findings provide useful insights for method design of flash crowds.