Abstract: An improved packet delay estimation method results from an indirect measurement method that employs an estimation method based on sampled lengths of queues and measured traffic. A complex system of queues sharing the same constant rate server and driven by a non-trivial scheduling scheme is split into logical clusters of queues. The method estimates packet delay based on a two-level approximation. On the first level of approximation, effects of the fact that the server is shared by other clusters is approximated by an equivalent reduction of the service rate. On the second level of approximation, individual queue lengths within a cluster are sampled and, based on the knowledge of the scheduling discipline, used to obtain the upper and lower bounds on the occupancy of the equivalent FIFO buffer. Estimate of the delay is found based on the effective buffer occupancy, and the effective service capacity.

Abstract: A simple and efficient method for packet loss measurement in high-speed switches is achieved by identifying traffic flows in the ingress side of switches, and measuring packets losses for the identifies flows on the egress side of the switches. Specifically, on the ingress side of the switch, the last packet of each fixed size incoming packet block is marked to make it distinct from other packets in the block. On the egress side, the packet flow is monitored and the packets are counted using the block size as a counting modulo. The value of this counter at the receipt of a marked packets is used to update the number of packets that do not make their way across the switch. The method is applicable to multicast as well as to unicast packets. This method is simple in that it only requires two marking counters per distinct flow, one at ingress and one at egress, and two other counters at the egress per each measured flow.