Data Sets for Time-Series Analysis

The format of these time-series is a sequence of bins, one bin per row, with the first number corresponding to the number of IP bytes in the bin, and the second number corresponding to the number of IP packets in the bin. Bins have a fix duration of either 1 or 10 milliseconds (file extensions 1ms and 10ms respectively). For example,

     84737        220
     67157        218

Datasets:

• Abilene-I (10-millisecond bins only)
  • A description of this data set can be found at the NLANR PMA Long Traces Archive.
• UNC02
  • Data set collected at UNC's main link (inbound direction only during April 2002.
  • Packet and bytes count time-series were prepared for only 20 traces (out of a total of 64 traces).
• UNC03
  • First data set collected at UNC's main link (inbound direction only) during April 2003.
• UNC03-bidir (10-millisecond bins only)
  • Second data set collected at UNC's main link during April 2003.
  • em0 and em1 correspond to outbound UNC and inbound UNC respectively.
  • The 10-ms. bins are not synchronized.

The UNC dataset are described in:

Cheolwoo Park, F. Hernández-Campos, J. S. Marron, and F. D. Smith.
Long-Range Dependence in a Changing Traffic Mix.
Computer Networks, vol. 48/3, pp. 401 ? 422, June, 2005.

Please acknowledge the use of these dataset in your work.