[OpenAFS] Questions about an AFS paper
Fri, 07 Apr 2006 00:44:00 -0400
I just finished reading "An Empirical Study of a Wide-Area
Distributed File System" by Spasojevic and Satyanarayanan. As many of
you know, this is an old paper (published in May 1996 in the ACM).
Spasojevic and Satyanarayanan turned to the xstat facility in order
to measure AFS's performance. The test lasted two 12-week periods (6
Here are my questions:
* Has there been a similar data collection effort over the last
* If there were to be a study, what facility should be used?
xstat? <xstat's replacement here>?
* If there were to be a study, would the community be willing to
participate in such a study?
* Should the community create/establish a set of
processes/procedures by which AFS performance/usage' patterns could be
monitored 'more often' (every 5 years instead of every 10)?
* Would the community gain anything from a study like that?
* Why did they used xstat facility to query remote AFS cells,
instead creating a small xstat program, distributed out to anyone who
would wanted to participate on the study, and have the program dump
performance data to the AFS namespace
(/afs/remote-cell.com/xstat-data-date), and then just crawl the
namespace? I might be missing something obvious, and if so, please just
let me know, as I have never used xstat, and I am just starting with AFS.
Thanks in advance,