Statistical Analysis of Packet Delay Time and Variation on the Internet

Delay is a crucial factor on the Internet especially for real-time applications like video or audio. The best-effort service model doesn't provide essential assistance to these applications because of the unstable queuing delays. In this work, we present the result of a measurement study designed to understand packet delay time and variation on the Internet. Measurements are performed between some vantage points in different countries using PlanetLab. Some statistical properties of the collected data were examined. It is inferred from the results that delay values are mostly appropriate for implementing real-time applications. Besides, delay over a path is stable in general while it has outliers from time to time. It means that distribution is right skewed. Values are clustered around the mode value and not dispersed too much.

I. Introduction

Multimedia applications have been getting more attention from Internet users and at the same time access speeds have increased. These applications have minimum bandwidth, loss rate, and delay requirements. The influence of packet delays in real-time applications can be reduced very little by using a buffer on the receiver side. Because streaming data cannot be buffered at end hosts long time. As a result delay becomes a crucial factor that affects the quality of such applications. The Internet is using the best-effort service model in which no guarantee is provided for packet delivery time. In default of such a guarantee of QoS (quality of service) in the Internet’s original design, researchers developed some methods, such as QoS, CDN (Content Delivery Network), and overlay techniques to improve the perceived quality for users.