[smokeping-users] Submillisecond RTT measurement accuracy

Russell Dwarshuis rjd at merit.edu
Mon Jan 18 17:04:58 CET 2010

For long-term reports I wrote a program using RRDs::fetch on the .rrd
files created by smokeping that analyzes all the pings in one shot.

Make sure your .rrd is sized correctly.  The stock row size will not
hold a month of 5 minute samples with step=1.  I set mine to:

# 110 days of all samples
AVERAGE  0.5   1  31680
# 400 days of every-2-hours consolidated data
AVERAGE  0.5  24  4800
    MIN  0.5  24  4800
    MAX  0.5  24  4800

You'll have to remove all your .rrd files if you decide to resize or
find a utility to convert them.

You'll also have to think of how to exclude data during maintenance
windows.  That will probably be a bigger challenge.

-Russell Dwarshuis

----- Original Message -----
From: "Thomas Klein" <smokeping at thomas-klein.de>
To: "Tobias Oetiker" <tobi at oetiker.ch>
Cc: smokeping-users at lists.oetiker.ch
Sent: Monday, January 18, 2010 7:38:52 AM GMT -05:00 US/Canada Eastern
Subject: Re: [smokeping-users] Submillisecond RTT measurement accuracy

Hi Tobias,

i have now setup smokeping on an ubuntu server and some basic icmp rtt 
measurements do work well.

Now i am wondering how to implement reporting for the SLO based on that. 
Maybe i can use the alerts and an external programm to log an entry for 
every treshold violation (median rtt > X ms) and report on the number of 
violations during a time interval. Not sure how an alert pattern must 
look for that.

Any other hints how to implement a monthly report showing if the median 
rtt was more than 10% of the measurement time above a given threshold?


Tobias Oetiker wrote:
> Hi Thomas,
> Today Thomas Klein wrote:
>> Hello,
>> i have to setup LAN measurements and reporting for a 0.9 ms RTT Service
>> Level Objective and try to investigate the Smokeping capabilities to do
>> that. Anybody on the list who has done something similar already or can
>> give me some hints about the accuracy i may can achieve with Smokeping?
> smokeping uses rrdtool internally for data storage, so there is no
> fundamental problem with even smaler numbers ... at worst it is a
> presentation issue ...
> the measurement happens with external tools anyway, so this again
> is not something which smokeping will influence ...
>> I am also wondering what method will be best, icmp ping, TCP syn, UDP
>> echo? Hints are welcome.
> this I don't know ...
> cheers
> tobi

smokeping-users mailing list
smokeping-users at lists.oetiker.ch

More information about the smokeping-users mailing list