We consider congestion control in peer-to-peer distributed systems. The problem can be reduced to the following scenario: Consider a set of peers (called clients in this paper) that want to send messages to a fixed common peer (called server in this paper). We assume that each client sends a message with probability and the server has a capacity of , i.e., it can recieve at most messages per round and excess messages are dropped. The server can modify these probabilities when clients send messages. Ideally, we wish to converge to a state with and for all . We propose a loosely self-stabilizing protocol with a slightly relaxed legimate state. Our protocol lets the system converge from any initial state to a state where and . This property is then maintained for rounds in expectation. In particular, the initial client probabilities and server variables are not necessarily well-defined, i.e., they may have arbitrary values. Our protocol uses only bits of memory where is length of node identifers, making it very lightweight. Finally we state a lower bound on the convergence time an see that our protocol performs asymptotically optimal (up to some polylogarithmic factor).
View on arXiv