The current PMA measurement design, which utilizes a collection of independent monitors, includes more than 21 passive monitors. These monitors take regular measurements on a variety of networks, ranging from FDDI to OC12 and from ATM to POS. Traces are collected, post-processed, encoded (anonymized), and then made publicly available on the NLANR PMA website for use by other network researchers, as well as students. The passive header trace data provides the means to study workload profiles for a number of measurement points in high speed environments, soon up to OC48, with work under way for OC192 packet tracing capability.
We have developed a number of tools to handle this data, all of which are available on the website. The tools can be classified as analysis or selection tools.
Analysis tools are usually programs which take our traces, or associated data, and either generate the data in a different format, or produce a human readable presentation of statistical data gathered from the traces.
Selection tools are predominantly Web based and are intended to assist researchers in finding suitable traces for their analysis needs.
To improve performance, all of the analysis tools are being rewritten in C (previously available only as Perl scripts). Three scripts have been rewritten successfully as of March 2002, with more underway. The initial performance increases have been quite promising. The tsh2ta program is now executing as much as 10 to 30 times faster than the equivalent Perl script (using equivalent hardware): 30 times faster on large traces, small traces are now 15- to 20-fold faster, with most runs experiencing a 10-fold increase. The reimplementation of ta2sum has resulted in a three- to four-fold increase in performance; while modest when compared to the gain in efficiency for tsh2ta, this is still significant.