Repository of Availability Traces

Maintained by Brighten Godfrey

This package includes, in a common file format, a distilled form of the data from a number of studies of machine availability and failure. The goal is to build a repository of availability data to make it easy for distributed systems researchers to obtain, use, and compare the data sets. Currently, the package includes traces of PlanetLab (Stribling 2005), web servers (Bakkaloglu et al 2002), corporate PCs (Bolosky et al 2000), DNS servers (Pang et al 2004), Skype superpeers (Guha et al 2006), and KAD peers (Steiner et al 2007).

Get more details: readme.pdf. The readme also includes a handy list of availability datasets beyond what is in this repository.


Download the collection: availability-0.3.tar.bz2 (4 MB)

New! steiner-kad.avt.bz2 (29 MB), a trace of a 179-day crawl of over 400,000 peers on KAD by Moritz Steiner, Taoufik En-Najjary, and Ernst W. Biersack, converted to .avt format by Lluís Pàmies i Juárez. Since it's much larger, this file is not included in the main collection.

List of availability datasets

Here's a woefully incomplete catalog of availability datasets, or papers that used such datasets, in reverse chronological order.

Included in this repository

Not included in this repository

Related projects