[rrd-users] average of 95th through 80th percentiles, in a sense

[Marius Rugan]

It seems to me it's a 95 percentile (top 5% of the samples are excluded)
combined
with an 85 percentile out of the remaining 95.
IMO think should be computing out of the MAX samples from the rrd and
consolidate with an average (how this can be done ?)
I'm also interested in the topic so i'll watch out.
Nice topic.

On Wed, Jun 18, 2008 at 7:28 PM, Ryan goldberg <
rgoldber-rrdtool at bitburst.com> wrote:

>
> Our upstream does 95th percentile billing in a weird way.  Quote:
>
> "Blahblah Telecom monitoring system takes and records a sample every five
> minutes for both the in and out going data stream.  At the end of the
> month,
> the samples are ranked and the top 5% of the samples are excluded and the
> next
> highest 15% recorded samples are used to compute the monthly sustained
> usage.
>  If the monthly average is larger than the base commitment level, the base
> commitment level is subtracted from the monthly average.  The result is
> used
> to bill the burstable usage.  The burstable useage rate is $blah/mps"
>
> Can someone point me in the right direction so that can I produce this
> number
> with rrdtool?  I have 60 days of 5 minute samples...
>
> TIA-
> Ryan
>
> _______________________________________________
> rrd-users mailing list
> rrd-users at lists.oetiker.ch
> https://lists.oetiker.ch/cgi-bin/listinfo/rrd-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.oetiker.ch/pipermail/rrd-users/attachments/20080619/1c2b39c0/attachment.html