[rrd-developers] patch for RRDtool support for aberrant behavior detection

[Jake Brutlag]

RRD Developers and Tobi Oetiker,

A few weeks ago, I introduced this mailing list to a development project
here at WebTV to integrate algorithms into RRDtool for application to
aberrant behavior detection in time series.

I am pleased to announce that a code patch is now available for the current
rrdtool-1.0.25.tar.gz distribution
at http://cricket.sourceforge.net/aberrant/rrd-1.0.25.aberrant.patch.gz. The
patch will update source code as well as the POD documentation files. My
experiences with the GNU patch program are limited, but here are the
instructions that worked for me:
1. Start with a clean install of rrdtool-1.0.25. That is, download
rrdtool-1.0.25.tar.gz and extract the tar archive, then extract the
directory structure and files from the tar archive. Don't run configure yet.
2. Uncompress rrd-1.0.25.aberrant.patch.gz into rrd-1.0.25.aberrant.patch
3. cd to the new directory (i.e. cd rrdtool-1.0.25)
4. Run the patch program with the -p1 flag:
patch -p1 < ../rrd-1.0.25.aberrant.patch
(assumes the patch is one directory up from the rrdtool install)
The -p1 flag tells patch to strip the first directory from the path of each
file listed in the patch file.
5. Backups of updated files are stored with *.orig extension.
6. Run configure, make, etc. as usual.

While the POD documentation should be accurate as to syntax and includes a
few examples, I encourage you to read the draft description of our
implementation at http://cricket.sourceforge.net/aberrant/rrd_hw.htm. The
document is primarily a discussion of implementation, not of the aberrant
behavior detection algorithm. This implementation touches many of the core C
files of RRDtool. At the same time, RRD file structure on disk is unchanged.
The enhanced tool will run with existing RRD files. This backwards
compatibility is essential, because we know our aberrant behavior detection
algorithm is only appropriate for a subset of time series. In some cases
simple thresholding (as Cricket provides) is sufficient. In others, the
processing cost of aberrant behavior detection is too high relative to the
potential benefit.

We are currently testing this software here at WebTV on our Solaris 2.5x
hosts. We haven't tested, compiled, or run the software on any other
platforms; and although everything should work (no new libraries were
added), this patch updates quite a few of the core RRDtool files.

A more complete description of the algorithms and methodology are
forthcoming in a LISA 2000 paper submission. I intend to post drafts of this
paper when it gets closer to final form. Support within Cricket for the
RRDtool enhancements is also forthcoming, and will checked into the
sourceforge repository development branch.

As always, your comments and feedback are welcome.

Sincerely,

Jake Brutlag
Network Analyst
Microsoft WebTV 

--
Unsubscribe mailto:rrd-developers-request at list.ee.ethz.ch?subject=unsubscribe
Help        mailto:rrd-developers-request at list.ee.ethz.ch?subject=help
Archive     http://www.ee.ethz.ch/~slist/rrd-developers
WebAdmin    http://www.ee.ethz.ch/~slist/lsg2.cgi