Weblog entry #244 for simonw
See all weblog entries made by simonw.
#244
mdev in iputils ping
Posted by simonw on Fri 25 Apr 2008 at 13:01
When you do a ping on recent versions of Debian you get a summary:
2 packets transmitted, 2 received, 0% packet loss, time 1015ms
rtt min/avg/max/mdev = 0.131/0.133/0.136/0.011 ms
I thought I'd check what "mdev" meant, after much searching, and eventually reading the source code (ping_common.c) I got the answer.
In the iputils implementation of ping "mdev" is the standard deviation of the round trip time.
Wikipedia has been corrected. I filed a bug report asking that the documentation and/or code be amended to make it more obvious.
It doesn't matter for most purposes which "deviation" is measured, as you'll usually just be interested in the relative, rather than absolute value of the number (assuming it is a good measure of deviation or variance), although if you were doing statistical checks on a network to establish if changes were significant or not it might be important to know.
I'm guessing the name "mdev" is a result of a translation issue. Some suggested it might be "mean deviation". However standard deviation can be calculated easily on an running basis, but I can't see how "mean deviation" could be calculated on such a basis (statisticians please tell me), so a ping that calculated "mean deviation" would have to store all the round trip times. Although I dare say it is possible to approximate the "mean deviation" in less resource intensive ways.
MacOS labels this "stddev" maybe here it does have a superior user interface.
2 packets transmitted, 2 received, 0% packet loss, time 1015ms
rtt min/avg/max/mdev = 0.131/0.133/0.136/0.011 ms
I thought I'd check what "mdev" meant, after much searching, and eventually reading the source code (ping_common.c) I got the answer.
In the iputils implementation of ping "mdev" is the standard deviation of the round trip time.
Wikipedia has been corrected. I filed a bug report asking that the documentation and/or code be amended to make it more obvious.
It doesn't matter for most purposes which "deviation" is measured, as you'll usually just be interested in the relative, rather than absolute value of the number (assuming it is a good measure of deviation or variance), although if you were doing statistical checks on a network to establish if changes were significant or not it might be important to know.
I'm guessing the name "mdev" is a result of a translation issue. Some suggested it might be "mean deviation". However standard deviation can be calculated easily on an running basis, but I can't see how "mean deviation" could be calculated on such a basis (statisticians please tell me), so a ping that calculated "mean deviation" would have to store all the round trip times. Although I dare say it is possible to approximate the "mean deviation" in less resource intensive ways.
MacOS labels this "stddev" maybe here it does have a superior user interface.