Link utilization

CCIE_2011

Hi everyone.

I was monitoring our network for a while using Solarwinds.

Our network is a quite big. Links in the dist to core are all fiber configured for 1Gbps.\


Solarwinds shows that we are only using, on avrg, 30Mbps of the 1Gbps. MAX recorded is 50Mbps. links Utilization are less than 5%.


is this a normal case for an enterprise network?
are we under utilizing our links ?
Is this a normal case for a health network?

Plus some rumors says that we will upgrade the network soon to be full 10Gbps in the core and dist area. If we use only less than 5% out of 1Gbps, I believe it is sufficient to keep it as is.

Thanks in advance for your comments.

networker050184

Its normal if you are only utilizing that much traffic. If you think you are using more than you can dig a little deeper to make sure solarwinds is reporting correctly.

On the subject of upgrading to 10G, we just upgraded our lab to a 10G connection that they just had to have. The highest utilization on the link has been like 50 Mbs. It sounds cool that we have 10G though

This is what happens when people with no real clue make the decisions. Oh well, plenty of room for growth I guess....

CCIE_2011

networker050184 wrote: »

If you think you are using more than you can dig a little deeper to make sure solarwinds is reporting correctly.

Any clue on how to check if solarwinds reporting is correct?

networker050184

CCIE_2011 wrote: »

Any clue on how to check if solarwinds reporting is correct?

Go to your devices and look at the utilization during peak times. Or you could use another system like Cacti and make sure they report the same utilization back.

hypnotoad

My problem with Solarwinds was that the counters were not accurate -- I dont remember why, but the gigabit links were reported incorrectly. This may have had something to do with the scaling it used. I don't recall at the moment if it was using megabits or bytes or 32 or 64 bit integers or dividing by 10 or 100 to scale the charts, but it wasn't very fun.

We tested the link by maxing it out, verified our results with a couple other methods, and compared those results to solarwinds.

sides14

I would have to agree. While we never verified the stats with another system, Solarwinds did report less BW usage than actual usage. We had 8 T1s in an IMA bundle that ran between Phoenix and Tucson via three microwave hops to a remote office (15 employees). The users at the remote office reported slow network speeds yet the senior admin always used Solarwinds to verify traffic.

CCIE_2011

sides14 wrote: »

I would have to agree. While we never verified the stats with another system, Solarwinds did report less BW usage than actual usage. We had 8 T1s in an IMA bundle that ran between Phoenix and Tucson via three microwave hops to a remote office (15 employees). The users at the remote office reported slow network speeds yet the senior admin always used Solarwinds to verify traffic.

Thanks for sharing for that info dude... we are about to convince managements to forget about solarwinds package

tiersten

hypnotoad wrote: »

My problem with Solarwinds was that the counters were not accurate -- I dont remember why, but the gigabit links were reported incorrectly. This may have had something to do with the scaling it used. I don't recall at the moment if it was using megabits or bytes or 32 or 64 bit integers or dividing by 10 or 100 to scale the charts, but it wasn't very fun.

Sounds like problems with the SNMP counters wrapping.

Forsaken_GA

We've found it much easier to just write our own scripts that poll for SNMP with proper counters and graph it using RRDTool. Cacti is very good about letting you do what you need, but it just had a little too much bloat for us.

If you're primarily a windows shop, take a look at PRTG. We've played around with it, and had good results, but management ultimately decided they'd rather pay our developers to actually develop instead of read blogs all day

Lizano

Forsaken_GA wrote: »

We've found it much easier to just write our own scripts that poll for SNMP with proper counters and graph it using RRDTool. Cacti is very good about letting you do what you need, but it just had a little too much bloat for us.

If you're primarily a windows shop, take a look at PRTG. We've played around with it, and had good results, but management ultimately decided they'd rather pay our developers to actually develop instead of read blogs all day

I´ve played around a bit with PRTG as well, but I´d like to know something, it seems like adding the devices and the sensors is kind of time consuming, is there a way I can do that, like a templates where I just change the IP address?

I can´t do auto-discover, and adding lots of hosts, and then the sensors to the hosts, looks kind of painfull...

Forsaken_GA

That, I can't answer, I was using auto-discovery when I played around with it, and management ultimately decided that they wanted a custom coded solution for production. I personally prefer a combination of Nagios and Cacti for my monitoring, so that's what I have deployed in my lab.