A tale of two outages: A study of the Skype network in distress

Abstract

This work applies snack, a flow-based algorithm for detecting connections between clients and supernodes to the Skype network, to the study of two separate outages on the Skype network in August 2007 and December 2010. We examine both outages in detail, comparing them to each other, and to previous work on the first outage. We find that a simple metric derived from connection events per unique client as measured by snack acts as a proxy for the Skype network, and would have provided forewarning of the 2007 outage, and may have applicability as an indicator of general Skype network health. We also examine the faster onset of and recovery from the 2010 outage, discovering that the distribution of clients per supernode is a proxy measurement for network centralization, and detects the apparent present recovery strategy from such outages. These two metrics may be applied by any network operator to monitor the health of the Skype network.

Identifying Skype nodes by NetFlow-based graph analysis

“…Skype is analyzed from a rather different perspective in [15] where two Skype outages are analyzed and compared. One of the interesting points of this work is that Skype temporarily centralized its topology by the introduction of so-called megasupernodes, thus speeding up the network recovery.…”

Web based teleoperation architecture and H.264 video encoder

“…These software solutions are easier to setup, allow multiple logins and updated regularly by their respective manufacturers. Drawbacks of these systems include system down times [6], sudden decrease in quality if the server becomes busy or low bandwidth is detected.…”