Community discussions

MikroTik App
 
Skyler
just joined
Topic Author
Posts: 9
Joined: Fri Sep 19, 2008 5:21 pm

Graphing discrepencies

Fri Sep 19, 2008 5:59 pm

I just started using Dude, and like it. It still has several quirks, most of which can be worked around, but this one is an issue of accuracy. I am running 3.0RC3, and noticed that when I view the bandwidth utilization on several of the switches that I monitor, the graphing appears to be off when changing Scale. If you look at the graphs I attached below, the "hour" view you can see 3 spikes just in that hour that went between 12 and 16 MBPS. But when you look at the "day" view, the graph shows the highest peek to be less that 10 MBPS. Something doesn't seem right there.

My biggest issue in using the Dude, is the lack of documentation. I'm more than willing to look into things, but there are several things within the 2.2 and 3.0 client that I can't find anything on, and my time is very limited. Beyond that, I think this is a very useful tool with a lot of good work having been put into it.

Thanks,
SKyler
You do not have the required permissions to view the files attached to this post.
 
netrat
Member
Member
Posts: 403
Joined: Thu Jun 07, 2007 1:16 pm
Location: Virginia

Re: Graphing discrepencies

Fri Sep 19, 2008 6:11 pm

If you change the from hourly to daily you are going to change the look of the graph. What is the average throughput for 10:00?? About 10mbps right? This is what the daily scale is going to show you.

I have the same thing using cacti for graphing. See I have two spikes to 3M on the daily graph, but these are not going to show on the weekly graph.
You do not have the required permissions to view the files attached to this post.
 
User avatar
jorj
Member
Member
Posts: 398
Joined: Mon Mar 12, 2007 4:34 pm
Location: /dev/null

Re: Graphing discrepencies

Sat Sep 20, 2008 10:55 pm

Actually, besides changing the scale, you are changing the interval that the graph is showing, somehow the "granularity" of the graphic. If the daily graph will calculate a 5 minute average, the weekly will show let's say a half an hour average.
That should enlight you.
 
lebowski
Forum Guru
Forum Guru
Posts: 1618
Joined: Wed Aug 27, 2008 5:17 pm

Re: Graphing discrepencies

Mon Sep 22, 2008 6:10 pm

The averaging of graphs over time is the only way to reduce the stored datasize... For example keep 60 samples for 1 hour and keep 48 samples for 1 day and ~720 samples for 1 month. If you tried to keep 60 samples for every hour for 1 month that would be ~43000 samples. That is a lot of samples... Averaging samples is the only way old data takes up less space but it becomes less meaningful since the max is averaged out.

There are other threads about this and there are several ideas for improving graphing... I don't know what they plan to do but one of the easiest things would be: If they allowed a choice to save the MAX for a counter instead of the average.

Although storing the MAX is also misleading... a given time range will show the MAX bandwidth used for that sample, it would look like your using more bandwidth than you really are but you would have a better idea of when you are starting to run out of bandwidth on a given link by looking at long term graphing.

Some people install Cacti or NMIS to have a long term graphing but you still need a tool to show MAX utilization, Max utilization is automatically stored in the RRD database and you can pull MAX out of the RRD databases. I don't know Cacti well enough but you create dashboards that show MAX with Ddraw or select mbits for NMIS. The samples in cacti and NMIS are only collected every 5 minutes but you can reduce that. You can customize how much data is stored in an RRD and you can find out how with a little research. For example store 20160 samples for 2 weeks instead of 300 samples for 2 weeks if the data is older than 2 weeks average that down to say 20,160 samples for 6 months. No mater what you do you have to throw some data away... The Dude is great for instantanious understanding of the network utilization.

Care should be taken when collecting SNMP data since you can start affecting device performance if you collect too often or too much.

Here are example graphs from NMIS for 48 days with bits(average) and mbits(max) counters displayed...
average.JPG
max.JPG
You do not have the required permissions to view the files attached to this post.
 
User avatar
jorj
Member
Member
Posts: 398
Joined: Mon Mar 12, 2007 4:34 pm
Location: /dev/null

Re: Graphing discrepencies

Tue Sep 23, 2008 9:45 am

Graphing the max would a very bad idea, if it were to replace the average graphing as it is now.

Besides, even on MT graphs, it is shown the max value, for the graph average, in text, not graph.

What I would want, and it's been already 2 ( yes, TWO ) years since it's been asked, and not only by me, to password protect the web page to the Mikrotik in whole, and allow only access to graphs. Yes, I know it's not that hard. And maybe one day........... :(

And also would be nice to include graphs for MORE than a year...... ( yes, I have machines running unmodified from 2005........ noob..... :) )
 
lebowski
Forum Guru
Forum Guru
Posts: 1618
Joined: Wed Aug 27, 2008 5:17 pm

Re: Graphing discrepencies

Wed Sep 24, 2008 12:53 am

Yeah they should just let you select max or average and store both figures... They definatly have a nice product here and there are many directions they could take it I would just like to see continued bug cleanup before they add more features.
 
Skyler
just joined
Topic Author
Posts: 9
Joined: Fri Sep 19, 2008 5:21 pm

Re: Graphing discrepencies

Thu Sep 25, 2008 7:50 pm

Thanks to everyone that replied. I did a bit of homework on this end (in between fires) and have a better understanding of what it being presented. It would be beneficial to have some means of showing "true" bandwidth utilization beyond the scope of the last hour. I occasionally have to look at issues regarding bandwidth utilization, and a true bandwidth graph for up to 24 hours would be extraordinarily useful.

Skyler
 
User avatar
normis
MikroTik Support
MikroTik Support
Posts: 24749
Joined: Fri May 28, 2004 11:04 am
Location: Riga, Latvia

Re: Graphing discrepencies

Fri Sep 26, 2008 12:10 pm

password protect is hard, given only the abilities that the MT webserver has. you can have a supersimple password, but anyone with some basic web knowledge will be able to bypass it.

ok, we could move the graphs behind the same authentication that is used for the webcfg ...
No answer to your question? How to write posts
 
User avatar
jorj
Member
Member
Posts: 398
Joined: Mon Mar 12, 2007 4:34 pm
Location: /dev/null

Re: Graphing discrepencies

Fri Sep 26, 2008 4:28 pm

password protect is hard, given only the abilities that the MT webserver has. you can have a supersimple password, but anyone with some basic web knowledge will be able to bypass it.

ok, we could move the graphs behind the same authentication that is used for the webcfg ...
Actually, you can look in the other treads treating this, What was wanted was a way of providing access ONLY to graphs, not the whole webpage shown when you access http://router_ip_address
Reason is this: Maybe there are people who won't bruteforce the webcfg if it's not shown. But there sure will be some that will do it if it;s shown. And there shure will be smarties that will use winbox if there's a link.
I am disabling web at all right now, for this reason.
For me it would be great to be able to allow everybody access to their graph, without access to anything else. I have clients with fixed public ip's, wich can watch when they want the use of their bandwidth.
 
lebowski
Forum Guru
Forum Guru
Posts: 1618
Joined: Wed Aug 27, 2008 5:17 pm

Re: Graphing discrepencies

Sat Sep 27, 2008 1:23 am

I don't know if they expose the counters in the XML that say how long to save data or if it is hard coded. A bit of digging in around in the XML backup should be easy to find out.

They should definitely let you tweak how long the collection stays before being averaged...

I would think data storage size would increase significantly if you are able to modify it from 1 hour to 24 hours... Although 400 megs would still be no big deal for a computer with a couple gigs. 400 megs is a guess... 20x more storage space maybe required and I have a 20 meg map.
 
bigd
just joined
Posts: 2
Joined: Mon Sep 15, 2008 5:18 pm

Re: Graphing discrepencies

Wed Oct 01, 2008 3:49 pm

It would be beneficial to have some means of showing "true" bandwidth utilization beyond the scope of the last hour. I occasionally have to look at issues regarding bandwidth utilization, and a true bandwidth graph for up to 24 hours would be extraordinarily useful.
They should definitely let you tweak how long the collection stays before being averaged...
I would think data storage size would increase significantly if you are able to modify it from 1 hour to 24 hours... Although 400 megs would still be no big deal for a computer with a couple gigs. 400 megs is a guess... 20x more storage space maybe required and I have a 20 meg map.
normis, this will greatly help in monitoring highly intermittent lines. If not 24 hours, 12 or 6 hours will do.

Who is online

Users browsing this forum: No registered users and 18 guests