An Internet paradigm and scheduling algorithm for guaranteeing end-to-end QoS
K. Mase, Shinya Nogami · 2003
In the proposed paradigm for guaranteeing end-to-end QoS, Internet service providers (ISPs) are responsible for guaranteeing the IP-level QoS for each QoS class. The maximum number of ISP networks a datagram can traverse is limited, and the end-to-end QoS target is uniformly allocated to each ISP. At each connection point (CP) between ISPs or between an ISP and an Internet service customer, bandwidth is contracted for each class for each direction of datagram traffic. The contracted bandwidth for incoming datagram traffic is split by the destination CP based on measurement; these split bandwidths are used to dimension the ISP network. Each router in an ISP network has queues for each flow (CP to CP) and each QoS class for each outgoing transmission link. A newly developed heuristic scheduling algorithm called "M/G/1 emulation" is used to control the maximum datagram delivery time for each queue for each class.