What is the best bandwidth monitoring technique?

[bhall]

I have a strange question on bandwidth monitoring. I have a Win2K network (300 clients and 11 servers), 12 cisco catalyst 3548XL switches, 2 cayalyst 4908G-L3 switches and 1 PIX 515.

I want to monitor the bandwidth of my network. My question is this. How do I get an accurate reading? Which device do I monitor?

I cannot see how I can get a true bandwidth reading by monitoring one device.

Do I monitor for example, VLAN1 on each switch, add them together then divide by the number of switches? Do I add the bandwith of each server and do the same?

Will monitoring JUST my two layer 2 switches (acting as my routers) give me an accurate reading for the entire network?

Any and all help appreciated.

[baileja]

My suggestion is get MRTG and monitor all devices and the ports that link to other infrastructure devices. MRTG is free and gives a very nice readable HTML output of bandwidth utilization for the SNMP manageable devices. My personal suggestion would be to monitor not the VLAN's but the ports between all the infrastructure devices. This will help point out bottlenecks. It is not bad to monitor VLAN's though, you can identify which ones are more utilized or have heavier traffic than others.

[bhall]

I found MRTG not long ago, also found PRTG, nice price for the Windows interface. I set my graphs for each server, each switch primary Vlan and each port on my routers.

My switches connect to my 4908G-L3 routers (layer 3 switches) through the GIG ports. Are you recommending to monitor those ports instead of the base VLAN? Will I have to monitor the same line on both sides?

My thought was that the VLAN would give me the best overall bandwidth for the switch.

Let's see if I have this right, I would monitor say, server 1 NIC and the switch port 0/1 it is connected to. The Router GIG port connected to my switch GIG port.

Does that sound about right? Thats alot of monitors going. (Alot even now). I'll see what impact this has on the network.

What do you think?

Thanks,

[baileja]

actually if you are monitoring the switchport that is connected to server 1 than there is no need to monitor the server nic for bandwidth (maybe if you wanted memory, processor, etc). I would definately monitor all uplink ports that go to another infrastructure device. It sounds lik a lot yes but you can put it in a very html page seperated by device. For example on mine, the first page is a list of devices, each with a picture of what kind of device (PIX, Cat6500, Cat4006, Cat3550, etc). When you click on that link it brings you to a summary of links on that device page (it has a link to the png 5 minute summare file for each link on that device). This give you a pretty nice view of all links on that device. If you want the particulars for one of those links you click on it and it brings you to the full page for that link (5 minute, day, week, and year).

[bhall]

I concure. It seems that for bandwidth, it really depends on what devices you use the most. After talking with the Directory of IT, we agreed that we do not need to monitor both ends. The uplink port will be good enough for our purposes. I am currently using PRTG (www.paessler.com), so I limited on the creativity of the web page. Great program though. It will send emails if I want. Thanks for your help.