[rrd-users] rrd graph inter/extrapolation of large U / NaN blocks

[Steve Shipway]

Here's another one.

4. Use LSL calculations to approximate the dataset as a line, and use that.
    VDEF:slope=x,LSLSLOPE
   VDEF:base=x,LSLINT
   CDEF:infill=slope,COUNT,*,base,+
   CDEF:newx=x,UN,infill,x,IF
  The downside is that you'll likely get a disconnect when switching between the real and infill data.  The upside is that it wont be a horizontal line and no need to store invalid data.

5. Use PREDICT function to fill in with averages of previous day's data.
   CDEF:infill=36000,1,1800,x,PREDICT
   CDEF:newx=x,UN,infill,x,IF
   This hops back 10hr (the size of your available-data window) and fills in with a 30min average of that data.

Of course, all of these infill ones have the downside that you're displaying made-up data for the unknown period...  it would be simple enough to make the RRD graph function display made-up data using a different colour of line, so that you can see which parts are true and which are cosmetic.

Steve

Steve Shipway
University of Auckland ITS
UNIX Systems Design Lead
s.shipway at auckland.ac.nz<mailto:s.shipway at auckland.ac.nz>
Ph: +64 9 373 7599 ext 86487

________________________________
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.oetiker.ch/pipermail/rrd-users/attachments/20130227/9f8f89b8/attachment.htm