I don't believe this is currently possible in a single pass, though it is a 
rather interesting idea...

The 'percent' RPN function, which calculates percentiles, is only available in 
a VDEF and so works on the entire visible dataset to give a single value.  You 
will also have the issue that you would need to extend the underlying RRAs to 
provide data at the best possible resolution for the entire graph width.  If 
you try to calculate a percentile using the 1-day granularity RRAs then you'll 
get artificially flattened values.

You could manage it by adding an additional manual calculation.

Create a new RRD file to hold the percentiles this should have 'last' RRAs, and 
a very large heartbeat and step (1 month for the step so as not to average, and 
with 'LAST' type RRAs so you can update it daily).  Run a periodic job (daily 
or monthly) to calculate the 95th percentile from the main RRD for the last 
calendar month, and store the value into the new RRD.

Then you can graph over both RRDs with the new one providing the 95th 
percentile lines.

Until RRDTool provides a percentile type CF for RRAs, or a percentile RPN 
function allowing a time window to be specified,  this is the best you can do.  
Maybe we can ask for these new features in RRDTool 2.x :)

Steve

Steve Shipway
University of Auckland ITS
UNIX Systems Design Lead
[email protected]
Ph: +64 9 373 7599 ext 86487


________________________________________
From: [email protected] 
[[email protected]] on behalf of 
KENNETH D MILLER II [[email protected]]
Sent: Thursday, 31 January 2013 10:30 a.m.
To: [email protected]
Subject: [rrd-users] 95th percentile per day,week,or month

Is there a way to graph a 95th percentile per day, week or month?  If I create 
a yearly interface graph, can i draw a flat line for each month's 95th 
percentile instead of one for the whole timespan?  I have added a modified png 
as an example here:
http://www.personal.psu.edu/kdm193/rrdtool/95thgraph.png


Thanks,
Ken

Systems Design Specialist
Network Development Engineering
Penn State University


Co-Chair Internet2 Performance Working Group
SuperComputing SCinet Measurement Team Member

_______________________________________________
rrd-users mailing list
[email protected]
https://lists.oetiker.ch/cgi-bin/listinfo/rrd-users

_______________________________________________
rrd-users mailing list
[email protected]
https://lists.oetiker.ch/cgi-bin/listinfo/rrd-users

Reply via email to