This content has been marked as final. Show 9 replies
Find out what the interface transmit and receive values are set to, and compare them with the actual interface. You can manually change the transmit and receive values in the interface details.
Ok, well that number would be correct for a 1MB port?!? I'm on a 100MB connection and the Tx/Rx settings are both 100000000. I'm having a hard time explaining to the boss how this is possible. Any additional help would be appreciated.
I just pushed for and got Solarwinds this year and I am quite surprised at interface utilizations over 100%. I'm talking about 700% to over 1000% utilization on 100 MB links for Cisco routers and other quipment such as a HP Unix server.
I need answers / A work- around/ resolution on this issue.
Solar winds are you paying attention here?????
Network Management Systems Engineer.
"Learn By Doing"
Network Performance Monitor discovers the port speed (bandwidth) from the router or switch.
If for some reason a 100MB switch interface is set to 1MB in the switch, then Network Performance Monitor will discover it as 1MB and calculate percent utilization base on that value. If you still want the switch to be set to 1MB, but want Network Performance Monitor to calculate the percent utilization based on the correct port speed. Right-click on the Interface from within Network Performance Monitor and select "Interface Details". Then manually change the discovered transmit and receive bandwidth to the values you want. This is the most common reason for unrealistic percent utilization.
The second most common reason is the SNMP agent on the remote device reporting bad values. There are many SNMP agents that will report good values in 32 bit counter mode, and report bad values in 64 bit counter mode. Are the traffic charts incorrect also. If so, then the remote SNMP agent may be reporting bad values. As an example, some Cisco 1200 Access Points report random values for the 64 bit packet counters. Switching NetPerfMon to 32 bit mode for these devices will work around the problem.
You can check the SNMP agent using SolarWinds MIB Browser or Real-Time Interface Monitor.
But... be sure to check the bandwidth settings on the Interface first. That's usually the culprit 95% of the time.
Many people will manually set the port speed of each interface to adjust the routing weights. This leads to incorrect port speed values.
Don, Thanks for the reply.
One the interfaces in question is on a Cisco 7206 router. The Fae interface is configured for 18MB leased. Now at differents times on any given day that interface may show 140% utilized.....that I do not comprehend.
I saw your explanantion, on a different thread, the same thing as it applies the Memory utilization on a server and the explanation made a lot made of sense to me. However, The routers ( Some Foundry and a few Ciscos) tend to show this problem even upon picking up the correct bandwith configuration from the interfaces.
Now are you saying that I need get on all of my serverfarm switches and configure a BW statement on every server port????? Thats a lot of work! Those ports are all fa/100MB full duplex with 100MB full-duplex on the server NICS config.
About the 32/64 counters, that was left at the default 32 bit during discovery and all the nodes I looked at all have the 64 bit checkbox unchecked.
Network Management Systems Engineer.
"Learn By Doing"
I do NOT reccomend adding a Bandwidth statement to each interface. That's a lot of work and cause other problems if you use OSPF or EIGRP. But... if a Bandwidth statement exists, Orion will pick it up and use it when calculating percent utilization of an interface. You can easily override any interface bandwidth setting from the Orion System Manager. Right click on the Interface and select "Interface Details". The Transmit and Receive bandwidths are at the top.
I have also seen Fast Ethernet interfaces on VIP cards report incorrect traffic stats if the statistics collection is set too short. If you set the statistics collection from 1 to 3 minutes from example.
The SNMP values in the router are only updated every 15 to 20 seconds (sometimes longer). So.. each time the SNMP counter are updated, 15 seconds worth of traffic are added to the snmp counter in the router. This is a statistical sampling issue.
Let me see if I can draw a picture of this. The following lines represent a time line. 'P' is an Orion statistics Poll and 'U' is the snmp counter update in the router.
As you can see, each Orion poll usually collects two counter updates 'U' during each poll. But... sometimes, you can collect up to three counter updates 'U' with a single poll. Now, this is usually not a problem unless the router holds on to too much data before updating the counter (which happens with VIP cards).
Increasing the statistics collection from 1 or 2 minutes to 6 to 8 minutes will fix this problem. Also, matching the statistics collection to a multiple of the counter update times can also help. You can tell how long the counter updates happen by using MIB Browser and contiously right-clicking on a value and selecting 'refresh'. So... if the counters are updating every 20 seconds, set the statistics collection to 400 seconds.
As I said before. This has only been an issue on interfaces that do not update their counters in real-time. Most interfaces update in real-time.
OK, Let's look at this from another angle.
What is the actual traffic rate you are seeing on the interface?
If the interface is a fast-ethernet 100Mbps and the traffic graph shows 140mbps, THEN you have a problem!
We have e100 WAN links with 30Mbps ATM PVC's on our Backbone routers.
I leave the Bandwidth setting on the router at 100Mbps & just set the interface speed to 30Mbps in Solarwinds.
In some cases we can oversubscribe & see utilization of %140, which is perfectly normal, but if you have the NetPerfMon bandwidth set to the actual bandwidth of the interface, then you should never see this.
Very Impressed with Engineers Edition Version 7.0 but have a similar issue with Bandwidth Utilisation on several 3Com switches, e.g,
Alert: Receive Percent Utilization of Raebarn_sever_room_4400_stack-RMON Port 26 on unit 1 · RMON Port 26 on unit 1 is 338 %.
Current traffic load of this interface is Received : 338 Mbps
Transmitted : 885 Mbps
This is a 3Com 4400 48 port 10/100 switch.
Following your combined advice on Net Perf MOn:
Interface Transmit and receive set to 100000000
SNMP set to 32 bit mode and the traffic charts are incorrect
Using MIB browser and selecting "refresh" on a cell updates instantly
SNMP statistics set to 9 minutes and Poll at 120 seconds.
Net Perf Mon helped resolve a helpdesk call for why server access across a 2Mbps WAN circuit was slow, large file transfer, but would like to monitor the end stations. Getting hundreds of theses email alerts from different 3Com sites so difficult to see the real data.
Any feedback appreciated
Network Development Technician
Portsmouth Hospitals NHS Trust