This paper analyzes end-to-end packet delay and loss behavior in the Internet using measured round trip delays of UDP probe packets sent at regular intervals. By varying the interval between probe packets, the study examines the Internet's load over different time scales, ranging from milliseconds to minutes. The results align with previous findings from simulation and experimental studies. The Internet's traffic is characterized as a mix of bulk traffic with larger packets and interactive traffic with smaller packets. The study observes packet compression (clustering) and rapid fluctuations in queueing delays. Probe packet losses are largely random unless the probe traffic heavily uses available bandwidth. The paper describes data collection using a UDP echo tool, which measures round trip delays at regular intervals. The data is analyzed to understand packet delay and loss characteristics. The results show that the delay distribution for all paths is best modeled by a constant plus gamma distribution, with parameters depending on the path and time of day. A spectral analysis reveals a diurnal cycle, indicating a base congestion level that changes slowly over time. Packet losses and reorderings are positively correlated with delay statistics. The study also examines the behavior of end-to-end round trip delays over shorter time scales. It finds that probe packets accumulate behind large Internet packets, leading to probe compression. The analysis of packet delay shows that when the interval between probe packets is small, queueing delays are more variable, and when the interval is large, delays are more evenly distributed. The results suggest that probe losses are random unless the probe traffic heavily uses available bandwidth. The paper concludes that the Internet's end-to-end performance remains an area requiring further research. The study provides insights into the behavior of packet delay and loss, which are essential for designing network algorithms, buffer sizing, and emerging applications like audio and video. The findings also highlight the importance of studying network dynamics for effective control mechanisms.This paper analyzes end-to-end packet delay and loss behavior in the Internet using measured round trip delays of UDP probe packets sent at regular intervals. By varying the interval between probe packets, the study examines the Internet's load over different time scales, ranging from milliseconds to minutes. The results align with previous findings from simulation and experimental studies. The Internet's traffic is characterized as a mix of bulk traffic with larger packets and interactive traffic with smaller packets. The study observes packet compression (clustering) and rapid fluctuations in queueing delays. Probe packet losses are largely random unless the probe traffic heavily uses available bandwidth. The paper describes data collection using a UDP echo tool, which measures round trip delays at regular intervals. The data is analyzed to understand packet delay and loss characteristics. The results show that the delay distribution for all paths is best modeled by a constant plus gamma distribution, with parameters depending on the path and time of day. A spectral analysis reveals a diurnal cycle, indicating a base congestion level that changes slowly over time. Packet losses and reorderings are positively correlated with delay statistics. The study also examines the behavior of end-to-end round trip delays over shorter time scales. It finds that probe packets accumulate behind large Internet packets, leading to probe compression. The analysis of packet delay shows that when the interval between probe packets is small, queueing delays are more variable, and when the interval is large, delays are more evenly distributed. The results suggest that probe losses are random unless the probe traffic heavily uses available bandwidth. The paper concludes that the Internet's end-to-end performance remains an area requiring further research. The study provides insights into the behavior of packet delay and loss, which are essential for designing network algorithms, buffer sizing, and emerging applications like audio and video. The findings also highlight the importance of studying network dynamics for effective control mechanisms.