Network Health Monitoring Program

Any advice for me on a simple to use program that would monitor the health (avail bandwidth, load, retransmissions, etc.) of a small network?

Big Network eth0
-- --------
| 1|----| switch |
-- --------
| |
-- | |
| 2|------ |
-- |
-- |
| 3|--------

Simplified diagram shows what the configuration is. If I run Ethereal on Box 1 and listen on eth1, I will be able to log all traffic going from any machie on the subnet to/from the big network, as well as from any machine on the subnet to/from Box 1. However I will miss the data between Box 2 and Box 3 since the switch will keep that data away from Box 1.

Okay, option 1: Find another tool that will log and analyze data like Ethereal, only set it up to somehow monitor the health of the hardware switch.

Option 2: Run another network monitor program on Box 2 (don't think I can use Ethereal, since Boxes 2 and 3 are QNX while Box 1 is Linux) and somehow line the data up with what was acquired on Box 1.

Option 3: Just assume that whatever bandwith is left over after running Ethereal and logging it, is always available to Boxes 2 and 3.

I really want to see some graphs and summaries of all the data that is being handled by the switch, and not just Box 1.

Any advice? Thanks in advance.


Re:Network Health Monitoring Program

If you switch is Cisco catalyst you can SPAN or MONITOR a port, group of ports or a VLAN to another
port. Not familiar with HP or 3Com but I would guess if the switch is a "manageable" switch capable of
VLANs then it would have a similiar feature. You could run tcpdump on Machine 2 and 3 and then read
the capture files with ethereal, won't be real-time but better than nothing. Addtionally there is a tool
called ettercap that may help.

Todd Magers

Re:Network Health Monitoring Program

Some switches can be configured to 'broadcast' all traffic out one port for such a monitoring purpose. Have to check the specs. on the switch to tell if it will do this. I believe the hp switches will, but some of the cheaper ones may not.