Explanations on RTTsd
-
Hi, sorry for the stupid question, what is the meaning of RTTsd , It is a new entry in pfsense 2.3 :)
Fabio
-
RTT Standard Deviation - a measure of how spread out the RTT is over the calculation period. The standard deviation is reported by other latency measuring tools, such as ping and mtr, though these tools usually calculate it on a 'whole run' basis, whereas dpinger uses a rolling calculation period.
On my VDSL2 connection, RTTsd is low - typically <2ms and often <0.5ms - it depends on how busy the link is.
On my cellular connection, RTTsd is huge - often >100ms. This connection is usually routed through a very busy 3G cell, with correspondingly poor latency figures. During any periods that the hardware manages to hold on to the weak 4G signal at my location, both RTT and RTTsd drop significantly.
-
Thanks David, now is a bit more clear. I ve a vdsl2 connection with (about) RTT 17 ms and RTTsd 4 ms.
to be more clear, both refer to the latency between pfsense and "cabinet on the street" but the difference is the calculation method ?
-
17ms is the mean (average) RTT. 4ms is a statistical measure called the standard deviation of how much the RTT is varying around that mean - the higher the RTTsd, the more the actual RTT figure has varied over the measurement period.
If the RTT figures follow a normal distribution - which is likely to be a reasonable approximation to the truth - then the actual RTT will fall within plus or minus two standard deviations of the mean 95.5% of the time. In other words, with a normally distributed RTT with a mean of 17ms and a standard deviation of 4ms, you would expect 191 of 200 RTT figures to fall between 17 - (4 x 2) = 11 ms and 17 + (4 x 2) = 25ms. Putting it another way, you have 95.5% confidence that the actual RTT will be between 11 and 25ms during the period in question.
At the moment, my VDSL2 connection has a RTT of 8.961ms and a RTTsd of 0.151ms. The low RTTsd indicates very low jitter on the link between pfSense and the ISP. This is not surprising as it's a 80/20 Mbit/s connection with less than 1 Mbit/s of traffic in each direction and there is little visible contention on either the infrastructure provider or ISP networks.
-
Both refer to the round trip time to the IP address being monitored. RTT is the average (mean) of the round trip time. RTTsd is the standard deviation of the round trip time, which gives you an idea of the variance in the measurements.
David included a link to the Wikipedia article on Standard Deviation. It's worth a read. From that link:
"The standard deviation is found by taking the square root of the average of the squared deviations of the values from their average value."
A dense, but accurate description. :)
-
Both refer to the round trip time to the IP address being monitored. RTT is the average (mean) of the round trip time. RTTsd is the standard deviation of the round trip time, which gives you an idea of the variance in the measurements.
There you have it from the horse's mouth. (If you didn't know, Fabioo, dennypage is the author of dpinger, the gateway monitoring program used in pfSense 2.3).
-
If the RTT figures follow a normal distribution - which is likely to be a reasonable approximation to the truth
That is a big assumption. The difference between my min and avg ping can be a magnitude less than my std-dev. You can't get a ping below your minimum, so you can get a uni-directional jitter resulting in what I think is more like a half-normal distribution.
-
If the RTT figures follow a normal distribution - which is likely to be a reasonable approximation to the truth
That is a big assumption. The difference between my min and avg ping can be a magnitude less than my std-dev. You can't get a ping below your minimum, so you can get a uni-directional jitter resulting in what I think is more like a half-normal distribution.
I agree - though I was trying to keep things simple to explain standard deviation to someone who didn't seem familiar with the concept. However, though you can't get a RTT below the physical minimum for the link, at higher latency the mean RTT shifts upwards and the RTT distribution spreads out around that elevated mean.
As you say, the real life distribution is likely to be a skew normal at the very least, with a strong skew towards high RTT and maybe skewed so heavily that the 'below the mean' end will be squashed almost flat. It might actually become an entirely different type of distribution.
To give an idea of the complexity that exists with packet delay variation, RFC 3393 is some light reading!
-
Thank you all, now I understand the functioning. and a double thanks to dennypage for dpinger :)
Fabio
