Stochastic modelling and analysis: a computational approach
Stochastic modelling and analysis: a computational approach
Analysis and simulation of a fair queueing algorithm
SIGCOMM '89 Symposium proceedings on Communications architectures & protocols
Linear optimization and extensions: theory and algorithms
Linear optimization and extensions: theory and algorithms
Random early detection gateways for congestion avoidance
IEEE/ACM Transactions on Networking (TON)
ISCC '00 Proceedings of the Fifth IEEE Symposium on Computers and Communications (ISCC 2000)
| Hi-index | 0.00 |
With the development of Internet and rapid growths of the network bandwidth, the quality of network transmission has a significant improvement. However, in the pluralistic network applications, the dramatic network bandwidth may be occupied by some of computers or users, such as ill-computer or improper peer-to-peer users. It affects the communication quality of service (QoS) of general network users or applications. How to deal with the unfair network bandwidth usage problem becomes a critical issue on network management. In general, the network bandwidth management can be realized by packet buffer management or output packet scheduling. In this paper, we formulate the packet admission control of buffer management as the Markov Decision Process (MDP) optimization problem. All of the received packets are classified into three types: real-time class, good-user class, and bad-user class. According to the QoS of packet class and traffic condition, the allowing policy derived by the MDP decides whether accept or discard the received packet for getting an optimal packet buffer management. With computer simulation, we demonstrate that the MDP-based buffer management can confine the buffer occupancy of bad-user class packet, and decrease the discarding rate of real-time class packet and good-user class packet significantly.