1: \begin{abstract}

2: 

3: The Transmission Control Protocol (TCP) is a Transport Protocol

4: used in the Internet. In \citep{O}, a more general class of

5: candidate Transport Protocols called ``protocols in the TCP

6: Paradigm'' is introduced. The long run objective of studying this

7: class is to find protocols with promising performance

8: characteristics. This paper studies Markov chain models derived

9: from protocols in the TCP Paradigm.

10: 

11: Protocols in the TCP Paradigm, as TCP, protect the network from

12: congestion by reducing the ``Congestion Window'' (the amount of

13: data allowed to be sent but not yet acknowledged) when there is

14: packet loss or packet marking, and increasing it when there is no

15: loss. When loss of different packets are assumed to be independent

16: events and the probability $p$ of loss is assumed to be constant,

17: the protocol gives rise to a Markov chain $\{W_n\}$, where $W_n$

18: is the size of the congestion window after the transmission of the

19: $n$-th packet.

20: 

21: For a wide class of such Markov chains, we prove weak convergence

22: results, after appropriate rescaling of time and space, as

23: $p\to0$. The limiting processes are defined by stochastic

24: differential equations. Depending on certain parameter values, the

25: stochastic differential equation can define an Ornstein-Uhlenbeck

26: process or can be driven by a Poisson process.

27: 

28: \end{abstract}

29: