S. Donnelly et al, "Passive Calibration of an Active Measurement System", to be presented at PAM2001.  Home/Search   Document Details and Download   Summary   Related Articles

....on the network, it is save to assume that the fluctuations are caused by random fluctuations in reading out the clock on either end. This suggest an error on each of the time stamps of the order of 0:006 s. Another approach is to calibrate our experiment with a similar setup. This is discussed in [15] resulting in a somewhat higher error estimate of 25 to 50 s. However, since network delays vary from O(100)s to O(1)s, these error estimates are small and the internal clock can be considered accurate enough for our purposes. VII. Software The top level Data Flow Diagram (DFD) of the software ....

....library uses the Berkeley Packet Filter software (BPF) BPF provides a raw interface to the data link layer and can access the packets as soon as they are released to the O S by the Ethernet interface. Our receiving program timestamps the packets at this moment. As will be shown in another paper [15], both the transmission latency (the time between sending a packet on a test box and seeing it pass by on the local Ethernet) as well as the reception latency (the time elapsed between a packet entering the box on the Ethernet cable and it being time stamped) are both of the order of 100 s. These ....

S. Donnelly et al, "Passive Calibration of an Active Measurement System", to be presented at PAM2001.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC