Reasons not to deploy RED

Abstract

In this paper we examine the benefits of random early detection (RED) by using a testbed made of two commercially available routers and up to 16 PCs to observe RED performance under a traffic load made of FTP transfers, together with HTTP traffic and non-responsive UDP flows. The main results we found were, first, that RED with small buffers does not improve significantly the performance of the network, in particular the overall throughput is smaller than with tail drop and the difference in delay is not significant. Second, parameter tuning in RED remains an inexact science, but has no big impact on the end-to-end performance. We argue that RED deployment is not straightforward, and we strongly recommend more research with realistic network settings to develop a full quantitative understanding of RED. Nevertheless, RED allows us to control the queue size with large buffers.