NEW Monitoring graph

[jdillard]

@NOYB:

Min: 0%
Avg: 1%
Max: 0%
See attached image.

I must have been daydreaming when that was being covered.  Could someone please tutor me before the test.

I forget why we decided to put that code in, but I just took it back out.

[NOYB]

I must have been daydreaming again when this was being covered too.  Could some please tutor me before the test?
How can the standard deviation be greater than the maximum value of the data set?

Probe Interval: 15000 (15 seconds)
Alert Interval: 15000 (15 seconds)
All others default

dpinger output:

loss: 0, delay: .0095890, stddev: .0079330
loss: 0, delay: .0091930, stddev: .0081340
loss: 0, delay: .0092940, stddev: .0080790
loss: 0, delay: .0097080, stddev: .0078550
loss: 0, delay: .1656320, stddev: .2779040
loss: 0, delay: .1651380, stddev: .2781920
loss: 0, delay: .1653530, stddev: .2780680
loss: 0, delay: .1652700, stddev: .2781160
loss: 0, delay: .0046320, stddev: .0013230
loss: 0, delay: .0055140, stddev: .0006520
loss: 0, delay: .0056070, stddev: .0007150
loss: 0, delay: .0052770, stddev: .0008860

Also see attached quality graph image.

2.3-BETA (i386)
built on Fri Mar 04 19:46:06 CST 2016


[jahonix]

I have a Nano install on a retired ALIX for testing 2.3
Problem is that I don't see any RRD graphs there (no menu entry to select), is that correct? I'm on a built from Feb 4th.
The old traffic graphs in the dashboard are still there, though.

[MikeV7896]

@jahonix:

I have a Nano install on a retired ALIX for testing 2.3
Problem is that I don't see any RRD graphs there (no menu entry to select), is that correct? I'm on a built from Feb 4th.
The old traffic graphs in the dashboard are still there, though.

Correct, the RRD graphs have been replaced with this new monitoring graph (Status menu > Monitoring). This graph is more dynamic, lets you turn off elements you don't want, and also allows you to show two sets of data at once. (compare interfaces, look at more data for one interface, etc.).

RRD is still collecting the data that is presented in this new graph, but it's not generating the images anymore.

[jahonix]

Thanks, even I found it now.
Was looking at "System" for "Monitoring", not at "Status" …

[biggsy]

WRT to NOYB's screenshot above, the units on some of the graphs and tables are a bit . . . strange.

[jahonix]

A) Is there a plan to convert the dashboard traffic graphs to the new "Interactive Graphs" as well?

B) two horizontal grids are "a bit" hard to work with. There should only be one grid at a time, scaling the axis accordingly. (see screenshot)

C) Black as text color was already mentioned not being perfect. Maybe we can have the grid in black to get better contrasts. Currently the white grid is the dominant color, making it harder than necessary to read the graphs itself.
The color scheme in NOYB's screenshot is OK, I'm probably just seeing a CSS issue.


[dennypage]

The delay number isn't the maximum value of the set, it is the mean average of the set. Dpinger reports average and standard deviation values against the entire set the same way ping does in its summary.

Standard deviation is the square root of the variance, which is based upon squares of the offset from the mean. Use of squares means that the result can be higher than the mean.You can read the lengthy discussion of standard deviation on Wikipedia, or you can use one of the on-line calculators to experiment.

Here is one that spells stuff out: http://www.mathsisfun.com/data/standard-deviation-calculator.html

Put "2,2,2,2,2,20" for the numbers and start experimenting from there.

Sorry about the cuteness of the page :)

@NOYB:

How can the standard deviation be greater than the maximum value of the data set?

[NOYB]

Yes I know std dev can be greater than the mean.  Because of the graph/table labels it seems to indicate that it is greater than the max.

So the graph/data table "delay" label is not very accurate.
Delay is really average delay.  Thus in the data table the min, avg, max, columns actually contain min avg delay, avg avg delay, max avg delay.

Labeling the std dev row as "stddev of delay" seems to indicate that it is the std dev of the row labeled "delay".  Which would be more accurately labeled "avg of delay".

[NOYB]

Sometimes the quality graph ends with zero.  I know this was corrected a while back but sometimes it still happens.

Kind of makes the minimum column useless.


[NOYB]

Changed to all default settings from:
Probe Interval: 15000
Alert Interval: 15000
All others default

And this was the result.  See attached image.

But running dpinger -f at command line reports std dev consistently between 700 and 800 microseconds.  Which is consistent with the gateways widget reporting.


[Clown]

Thanks, Monitoring works now.

There should be some rounding of the values, see the attached Picture.

Edit:
This happens just when the 'Timer period' is greater than 1day.

[Harvy66]

@dennypage:

The delay number isn't the maximum value of the set, it is the mean average of the set. Dpinger reports average and standard deviation values against the entire set the same way ping does in its summary.

Standard deviation is the square root of the variance, which is based upon squares of the offset from the mean. Use of squares means that the result can be higher than the mean.You can read the lengthy discussion of standard deviation on Wikipedia, or you can use one of the on-line calculators to experiment.

Here is one that spells stuff out: http://www.mathsisfun.com/data/standard-deviation-calculator.html

Put "2,2,2,2,2,20" for the numbers and start experimenting from there.

Sorry about the cuteness of the page :)

@NOYB:

How can the standard deviation be greater than the maximum value of the data set?

Should probably add a "*" and have a note of something like "*all reported values are aggregated over a rolling window". "Min" and "Max" are a bit deceiving when they'er averages themselves. Not in a bad way, but in a non-intuitive way that can cause confusion with some datasets.

[samoth]

Any reason why my graphs wouldn't have any data?  No matter what I pick in settings, I get 0 for values in the graph.

[grandrivers]

packet loss should be red in color

[NOYB]

Could we have more granularity than 5 minutes please?  Especially on the 1 hour graph.
The RRD has 60 data points an hour but the graph seem to display a granularity of only 12 data points per hour.

Makes for nice smooth lines that look pretty for consumer class.  Ideally a granularity selector based on the RRD data granularity of the selected graph time period would be nice.

Thanks.

[xbipin]

i have a full install where monitoring graphs work but i have another nanobsd but monitoring shows blank graphs there, is there some way to enable it because on this nanobsd install i never previously enabled RRD graphs

[jdillard]

@NOYB:

Yes I know std dev can be greater than the mean.  Because of the graph/table labels it seems to indicate that it is greater than the max.

So the graph/data table "delay" label is not very accurate.
Delay is really average delay.  Thus in the data table the min, avg, max, columns actually contain min avg delay, avg avg delay, max avg delay.

Labeling the std dev row as "stddev of delay" seems to indicate that it is the std dev of the row labeled "delay".  Which would be more accurately labeled "avg of delay".

So "delay" should be "avg delay" and "stddev of delay" should just be "stddev"?

[jdillard]

@grandrivers:

packet loss should be red in color

manually assigning colors is a bad idea for this use case. Possibly in the future, more specialized charts could get special treatment.

[dennypage]

I think this would be sufficiently clear.

@jdillard:

So "delay" should be "avg delay" and "stddev of delay" should just be "stddev"?