|
| | | Forum Newbie
Group: Forum Members
Last Login: 7/1/2009 8:26:53 AM
Posts: 4, Visits: 18 |
| Hi all, I discovered a funny thing. When looking at interface statistics of gigabit interfaces on a Cisco3750 the displayed graph is not correct. When I install another monitoring application like PRTG or Solarwinds on the Whatsup server the statistics are correct. I use Iperf to generate traffic so I know exactly how much traffic is generated. Anybody who has seen the same issue? To be sure that I didn't misconfigure anything I installed Whatsup Professional 2006 and PRTG from scratch and did the same tests. To let you see the difference see the attachments. Hope to hear from you... Erik PS: the dates in the graph are from last year but the situation still exists, even after I upgraded to Whatsup Gold V11.
|
| |
| | | Time Traveler
Group: Ipswitch Employees
Last Login: 8/14/2009 3:06:04 PM
Posts: 502, Visits: 1,302 |
| WUP 2006 does have a problem, where it displays an incorrect value (and extra zero) for the devices bandwidth (but I've been told it does not affect the calculations!), that has been fixed in WUG V11. WUGV11 also added the ability to enter a value for the Interface Speed, as some devices reported an incorrect value. To set the interface speed, go to the Device Properties, Performance Monitors and double click the desired monitor (Interface Utilization), the 'Collect data for:' setting must be set to Specific Interfaces and any that are checked, can then be selected and then when you click the Speed button, you can enable 'Specify Speed' and input the speed for the incoming and/or outgoing port. Daniel Donnelly |
| |
| | | Forum Newbie
Group: Forum Members
Last Login: 7/1/2009 8:26:53 AM
Posts: 4, Visits: 18 |
| Thanks for your reply. I will do another test with WUG V11 compared to PRTG and Solarwinds and will let you now. I knew about the interface speed setting and I am very pleased with it. Since we use an IP-VPN service all interfaces are ethernet NIC's allthough the IP-VPN services vary from 512 Kbps to 34 Mbps. With this setting you can have a much better picture of what is going on. Erik |
| |
| | | Forum Newbie
Group: Forum Members
Last Login: 10/8/2007 1:53:42 AM
Posts: 7, Visits: 8 |
| | Is it really fixed in WUG v11? We got the same Problem as you... we don't trust in the Performance Monitors, that WUG shows us... there are big differences, when changing the collection interval from 10 minutes to 1 minute. (see screenshot) I do not know if the values are right, when collecting with 1 minute... I think they are not a 100% correct! does anybody know, how WUG v11 creates the graphs? does it only collect the current value every minute (or every 10 minutes), and then it does calculate itself the average or the maximum??
|
| |
| | | 
Time Traveler
Group: WhatsUp Gold Expert
Last Login: 7/13/2009 12:55:00 PM
Posts: 1,593, Visits: 7,794 |
| In my experience, the Wug stats are correct. The only thing you have to keep in mind is that the sampling period *does* play a BIG role in statistics collection ! This does not apply only to Wup but to any other system also... Don't forget that the inOctets and outOctets snmp values are COUNTERS. So, what happens when you collect data ? Let's look at inOctets. To get them you have to do this : Sample at T0 = 1000, I'll call it St0 Sample at T1 = 8000, I'll call it St1 Number of octets received between T0 and T1 = 8000 - 1000 Octets / sec = (St1-St0)/(T1-T0) Now, given the "random" nature of network traffic, the higher the sample period, the bigger the approximation. Imagine for instance that you poll every 60secs. It is *possible*, and it does happen, that you have 1gbit/s of traffic during the first 30secs, then no traffic at all during the other 30secs. Your stat will give you... 500mbits/sec. Now if you sample every 30secs, you will have one sample at 1gbit, and the next sample at 0... Besides, based on the data retention settings you put in the Wup console, Wup will "cleanup" older records and resample them (raw data resampled to hourly data, then hourly data resampled to daily data). This process allows to keep the database size reasonable, but you do lose accuracy in the measurement. As always, remember what Winston Churchill said : "the only statistics you can trust are those you falsified yourself" 
Reading, writing and arithmetic - If you need to choose, please take option 1. |
| |
| | | Forum Newbie
Group: Forum Members
Last Login: 9/10/2007 5:34:41 PM
Posts: 4, Visits: 39 |
| We've been working on a similar issue where we were trying to figure out why values moving from one device didn't correspond as expected within the calculations or in the graphs. What we think is that the limit within the SNMP counter being too low for the volume of data for that time frame.
The limit for the SNMP counter is 4.294967296 GB. Depending on where in the counter sequence you begin recording the information you could max the counter and start over within the time frame and the calculation will be incorrect. What we've done is to shorten our time frames within the performance monitor and have fixed this issue. The inherent problem is that if your data stream is too high you could potentially flip the counter in less than a minute.
The attached graphs show from source PC to switch then to the server switch and to the destination server and we used 17.5 GB of data to standardize this test. The time durations we chose were 10 minutes and 1 minute respectively. Looking at the first segment in all the graphs you see that they are different when you move from one device to another. Then in the second segment there are closer similarities, but there are still minor differences.
|
| |
| | | Forum Newbie
Group: Forum Members
Last Login: 7/1/2009 8:26:53 AM
Posts: 4, Visits: 18 |
| Hi all,
It is fixed in WUG 12.4. I tested version 12.4 recently and discovered that the statistics are ok now. Issue solved.
I rest my case.
Cheerz, Erik!
|
| |
|
|