Network Bandwidth Analysis

[petben]

Hello, we have a network setup of mostly 1G Netgear GSxxx switches connected by fibre. I feel there is a case for a 10G backbone upgrade, but 10G switches are expensive, and we need to add some hard evidence from monitoring the current traffic, maybe we don't get anywhere near even 1G... Is there any tool/method anyone can recommend? There doesn't seem to be anything on the switch interfaces...

 

Thanks

[supportman]

You could try copying a file over the network switch and then on a small 10gb switch and compare the results.

[petben]

Thanks but we need to see in our current environment whether the 1G connection is currently maxing out.

[FN-GM]

We use Zabbix. It is free and allows us to monitor the switches via SNMP. We can see exactly how much bandwidth is used and when. it will produce it in nice graphs for you. I have attached an example - in this case it is an internet connection.

[Garacesh]

Another +1 for Zabbix. We use it to monitor incoming/outgoing traffic as well as UPS power levels/expected runtime.

[AlanD]

Hello, we have a network setup of mostly 1G Netgear GSxxx switches connected by fibre. I feel there is a case for a 10G backbone upgrade, but 10G switches are expensive, and we need to add some hard evidence from monitoring the current traffic, maybe we don't get anywhere near even 1G... Is there any tool/method anyone can recommend? There doesn't seem to be anything on the switch interfaces...

 

Thanks

 

Netgear do have a tool - NMS300 which is free for first couple of hundred devices. You could use google images to see the kind of stuff it will produce. You can also use spiceworks and other stuff with SNMP for which netgear has all the conifiguration files for their devices.

 

However...I would start by looking at the data coming from you servers using the windows task manager tool. My guess is that this will tell you that you are no where near even 1Gb/s even when you aggregate stuff from all the servers and add in the data through your firewall/filter. Unless you have SSDs in your server the random access data rates are likely to be quite limited especially - as is often the case cost savings have been made by using midline 7K2 drives rather than 15K SAS drives in the server.....which can have a significant effect on the datarate - because 'even if' (and to some extent 'even because') you have a large disk array (which might help sequential access) it won't help random access rates....as that will be determined by the access time for the last disk to respond in the array....which will be a full revolution for a large array.

 

The GSxxx switches all support Link aggregation (albeit static rather than dynamic for the oldest models)….so assuming your fibre installation has spare fibres (I would say only an idiot would install fibre runs without extra pairs..be3cuase it costs so little extra...but sometimes cost is everything) you can simply add a second fibre link between switches....or a copper one....and assign them to a LAG. This alone, would give you 2Gb/s and also give you redundancy should a fibre or fibre module fail.

 

I think I would check all the firmware is up to date before putting in the links....because some of the early firmware - as with most brands....could be buggy with LAGs and VLANs.

 

And...I would be getting a load of cheap £10 fibre modules from ebay/china because (a) the genuine netgear ones are eyewatering expensive (b) the genuine ones fail with painful regularity © the cheap ones work even better and never seem to fail.

 

...absolutely no point of course..in updating your backbone if you are not saturating the links coming from your servers though...and if you are...again I would be putting in a aggregated link from the server to the main switch as a first step.

 

I would also suggest looking at ebay for 10G netgear switches...something like GSM7352S is often under a £100 ...and although quite old...works fine. (OK...before I hear complaints...I know the fans are noisy....)

[pantscat]

iPerf might help?

Not sure about those Netgear GS models, but you might be able to get some port usage stats using some sort of SNMP monitoring tool?

[Boredguy]

We use Cacti to monitor all our netgear switches and never get near the limit on our 1Gb fibre, even with 90 stations hanging off the end of it.

[mavhc]

Installed Zabbix appliance VM, now just need time to read about SNMPv3 and getting it working with Zabbix.

 

Deployed the Windows agent to the computers

[pete]

Observium ( Observium ) or LibreNMS ( https://www.librenms.org/ ) can also do this trivially.

[ADMaster]

I've been using Librenms for a month now. It was very easy to compared to others. It really prefers host names over IP addresses though.

[ITGuyWestMidlands]

Netgear do have a tool - NMS300 which is free for first couple of hundred devices. You could use google images to see the kind of stuff it will produce. You can also use spiceworks and other stuff with SNMP for which netgear has all the conifiguration files for their devices.

 

However...I would start by looking at the data coming from you servers using the windows task manager tool. My guess is that this will tell you that you are no where near even 1Gb/s even when you aggregate stuff from all the servers and add in the data through your firewall/filter. Unless you have SSDs in your server the random access data rates are likely to be quite limited especially - as is often the case cost savings have been made by using midline 7K2 drives rather than 15K SAS drives in the server.....which can have a significant effect on the datarate - because 'even if' (and to some extent 'even because') you have a large disk array (which might help sequential access) it won't help random access rates....as that will be determined by the access time for the last disk to respond in the array....which will be a full revolution for a large array.

 

The GSxxx switches all support Link aggregation (albeit static rather than dynamic for the oldest models)….so assuming your fibre installation has spare fibres (I would say only an idiot would install fibre runs without extra pairs..be3cuase it costs so little extra...but sometimes cost is everything) you can simply add a second fibre link between switches....or a copper one....and assign them to a LAG. This alone, would give you 2Gb/s and also give you redundancy should a fibre or fibre module fail.

 

I think I would check all the firmware is up to date before putting in the links....because some of the early firmware - as with most brands....could be buggy with LAGs and VLANs.

 

And...I would be getting a load of cheap £10 fibre modules from ebay/china because (a) the genuine netgear ones are eyewatering expensive (b) the genuine ones fail with painful regularity © the cheap ones work even better and never seem to fail.

 

...absolutely no point of course..in updating your backbone if you are not saturating the links coming from your servers though...and if you are...again I would be putting in a aggregated link from the server to the main switch as a first step.

 

I would also suggest looking at ebay for 10G netgear switches...something like GSM7352S is often under a £100 ...and although quite old...works fine. (OK...before I hear complaints...I know the fans are noisy....)

This. It's great you are looking to monitor the links first. So much money is wasted on new 10G switches for no real reason.

 

NMS is what you need, assuming all your switches are Netgear.

[supportman]

I've been using Librenms for a month now. It was very easy to compared to others. It really prefers host names over IP addresses though.

 

Has anyone put it on Hyper-V? Just struggling getting the VM imported.

Edited January 14, 2019 by supportman

[PotNoodleTech]

Zabbix sounds interesting!

[mdrabble]

Just install Zabbix, just hoping I;ve set up the SNMP monitoring correctly!

 

Hopefully after an hour or so, I'm hoping to see some results, otherwise there may be a new thread about it

[PotNoodleTech]

Just install Zabbix, just hoping I;ve set up the SNMP monitoring correctly!

 

Hopefully after an hour or so, I'm hoping to see some results, otherwise there may be a new thread about it

 

Did you use the production version or the appliance version?

[mdrabble]

Did you use the production version or the appliance version?

 

I used the production version.

 

Created Ubuntu VM and in the Zabbix installation instruction are very easy to follow

 

Just had one issue with php time zone, where the Zabbix instructions mentioned a file to edit, which didnt work, so I ended up editting the php.ini with the correct timezone details.

[PotNoodleTech]

Ok I am trying out the HyperV appliance version as I am a Linux Luddite

[supportman]

Ok I am trying out the HyperV appliance version as I am a Linux Luddite

 

I tested but just got a black screen. No idea whats wrong.

[ADMaster]

LibreNMS doesn't have a pre built appliance for hyper-v, but yes I have it running in Hyper-v.

I use Ubuntu so this is the guide I used. I may have needed to edit the php.ini file for time zone too. I installed 3-4 NMS systems with in a week so don't recall for sure. Librenms is the one I liked best.

 

https://docs.librenms.org/Installation/Installation-Ubuntu-1804-Apache/

[mikeyd101]

I tested but just got a black screen. No idea whats wrong.

I started looking at this and ran it this issue too. You have to switch terminal sessions CTL + ALT + F2

 

Also if your using appliance version there are a couple of places I needed to change timezone to get my times correct.