A Measurement-based Study of End-to-End Internet Traffic Dynamics
نویسندگان
چکیده
We present findings from a multiple week study of Internet traffic dynamics using link throughput as the primary metric. Many traces were conducted through a four node, mesh-connected, “real world” Internet testbed. We generate different types of traffic (single TCP flows, multiple TCP flows, and single UDP flows) using a generic, multipurpose, sockets program. We discuss 1) link asymmetry for TCP and UDP traffic; 2) effects of different transfer sizes on throughput when using TCP and UDP; 3) link throughput when sending multiple simultaneous TCP flows, comparing their total throughput with a single TCP flow’s throughput; 4) link throughput achieved when using different transport protocols; and 5) the changing trend of diurnal traffic patterns. We find that there continues to be significant asymmetry between sending and receiving paths. Also, we show that TCP throughput is significantly affected by transfer size, however UDP traffic throughput is impervious to transfer size variations. We further find that larger (longer duration) TCP flows share bandwidth more fairly than smaller TCP flows; also, no significant throughput improvement was noted when using larger flows, however multiple small flows can achieve a significantly higher throughput than a single small flow. Finally, we show that UDP traffic always attains the available bandwidth, which is significantly higher than that attained by TCP traffic.
منابع مشابه
TEMB: Tool for End-to-End Measurement of Available Bandwidth
The explosive growth of the Internet has induced a need for developing tools to understand the composition and dynamics of the Internet traffic. Measurements of the various characteristics of a network provide insight into the state and performance of the network. The end-to-end available bandwidth metric can be used by various applications for various purposes like path selection, dynamic rout...
متن کاملOn Measuring the End-to-End Packet Delay and its Dynamics Modeling using System Identification
Understanding the end-to-end packet delay dynamics of the Internet is of crucial importance since it directly affects the QoS (Quality of Services) of realtime services, and it enables us to design an efficient congestion control mechanism. In this paper, we measure the end-to-end packet delay of the Internet, and build a model representing its dynamics using the system identification. We first...
متن کاملDynamics of Feedback-Induced Packet Delay in ISP Router-Level Topologies
Internet behavior is becoming more complex due to everchanging networking technologies and applications. Thus, understanding and controlling the complex behavior of the Internet are important for designing future networks. One of the complex behaviors of the Internet is traffic dynamics. Previous studies revealed that flow control in the transport layer affects the traffic dynamics of the Inter...
متن کاملEnd-to-End QoS Support for Adaptive Applications Over the Internet
In current end systems, multiple flexible, complex and distributed applications concurrently share and compete both end systems resources and transmission bandwidth of heterogeneous multi-protocol networks, especially the Internet. Our objective is to enable adaptation awareness in these applications to fully cope with the dynamics in resource availability over the heterogeneous Internet, as we...
متن کاملSelf-Similar Traffic and Network Dynamics
One of the most significant findings of traffic measurement studies over the last decade has been the observed self-similarity in packet network traffic. Subsequent research has focused on the origins of this self-similarity, and the network engineering significance of this phenomenon. This paper reviews what is currently known about network traffic self-similarity and its significance. We then...
متن کامل