A hop-by-hop flow controller for a virtual path

Adaptive Predictive Control of ATM Networks with Multiple Nodes

“…Tan [5] discussed the single-node problem, and obtained the locally stable PI controller; Ren [6] designed the PID controller in frequent domain, and controller parameters option, but the algorithm was not to suitable to the networks with variant users and time-varying delay; Benmohamed [7] proposed PD controller, and realized the local stability, however, the nonlinearity was not considered there; After that, Quet [8] presented a stable ∞ H control algorithm for multiple time-varying delay cases, and obtained the steady-state fairness, but no efforts on dynamical users and nonlinearity. Paralleling to the single-node case, Benmohamed [9] extended his single-node case to multiple-node ones [7] ; Zhang [10] designed a robust control algorithm for hop-by-hop networks; Biberovic [11] extended Quet [8] to multiple-bottlenecked networks. They all didn't consider the effects on time-varying delay, variant users and nonlinearities.…”

Rate-Based Congestion Controllers for High-Speed Computer Networks

“…In the case of window control, a maximum number of cells (or packets) is specified that a source can transmit, the window size thus limits the maximum number of cells, and hence the source throughput, that can be transmitted in a round-trip interval; while the rate-based control, for example, forward explicit congestion notification (FECN) and backward explicit congestion notification (BECN) (Yang and Reddy, 1995), regulates the source rate based on feedback information on the buffer occupancy in the switching node. These kinds of control strategies are very effective in conventional packet-switched networks and attracting increasing research interests (Kalarov and Ramamurthy, 1997;Zhang et al, 2000). Stability of closed-loop system is critical in any congestion control scheme due to the fact that, propagation delay encountered in high-speed networks may cause the controllers and the whole network to operate at an unstable point.…”

“…Mostly recently in Mascolo (2000) Smith's principle was applied in designing a control law for ABR input rates in ATM networks. Zhang et al (2000) proposed a hop-by-hop congestion controller designing method, in which system stability was met to the occasionally chosen controller. Such specially chosen controller guaranteed the requirement of system stability in this special occasion, but may not be able to meet other performance requirements of actual network in other circumstances, for example, to limit the duration of response time and maximize the throughput (Schwartz, 1996).…”