etherchannel on cisco 3548 switch?

sorry, heres some version info:


IOS (tm) C3500XL Software (C3500XL-C3H2S-M), Version 12.0(5)WC7, RELEASE SOFTWARE (fc1)

cisco WS-C3548-XL (PowerPC403) processor (revision 0x01) with 16384K/1024K bytes of memory.

i have a 3560 switch elsewhere on my network and have verified that the channel-group command exists on that one.

yeah it supports etherchannel..... how many links are you bonding?

Thanks
scott

Sorry- it's "port group" without the hyphen.

ive got a number of hp servers that have a teamed nic, and realized that they were setup as a team, but the switch was never configured for it.  i will probably only do 5 or 6 of the servers that get the heavier traffic.

port group is there.  thanks for that info.

can this be done with the servers online, or it going to interupt traffic at all?

you should not interupt traffic it should just pick up the links one at a time and bundle them into the port group.  if you are only having 2 ports per group then you should be fine but I would still do it off hours just incase.

Thanks
scott

Well... the problem is that the switch will only support Cisco's proprietary PAGP and not the IEE standard LACP. So it's unlikely that you can run your servers with port group on the 3524.

There are nic's out there Intel's I believe that did at one time offer this functionality.

Thanks
scott

from what i understand though is that you can force the pagp to 'on' and that works with non-compatible devices.  is that not true?

Not sure frankly, but I suspect there's more to it than that. PAGP does the negotiation and that can be turned off, but the devices still have to communicate port status via a protocol. If they don't line up correctly, problems occur.

this is also what i found on the link provided above:

...and on (unconditionally sets the port to channel and does not exchange PAgP frames). A connection between a PAgP-capable switch and a non-PAgP device requires the on mode to form a channel.

It should work but you are taking a chance that the host will not be able to reassemble the packet and will request it again.  I would set it up in a test enviroment checking for error's  I have only seen this done with Intel nic's and they were purpose built for this type of load balancing.

Thanks
Scott

the hp servers come with the software installed to team the nics together.  they dont seem to be using the full capacity with no configuration on the switch right now.  also because each nic has a different mac, the switch doesnt seem to know about the second nic on each server.  how else would i go about getting this team to work properly?

this is one of the nics from a dl380:
HP NC7782 Gigabit Server Adapter #2

it has an item listed: 'Network Load Balancing' which is described as "This component provides TCP/IP load balancing functionality".


the hp software has a number of options for load balancing.  the current mode is "Transmit Load Balancing with Fault Tolerance (TLB)".  this is with the switch having no configuration and the team mode on the server is set to auto.
other options:
-802.3ad Dynamic with Fault Tolerance
-Switch-assisted Load Balancing with Fault Tolerance (SLB)
-Transmit Load Balancing with Fault Tolerance (TLB)
-Transmit Load Balancing with Fault Tolerance and Preference Order
-Network Fault Tolerance Only (NFT)
-Network Fault Tolerance Only with Preference Order

i cant find anything that mentions PAgP.  does this mean that only the server can transmit with load balance since it cant communicate with the switch?  would a newer switch enable me to get better functionality?

A newer switch definitely would work better, because it will do the standard LACP. But as suggested, you can try what you have during a maintenance window or in a lab.

Not to mention the 3548XL's don't have best reputation for stability and performace...the are also EOL and EOS so even if you wanted to get support on them you could not.

Thanks
scott

i do have fiber links between switches throughout my network.  the servers arent pounded that hard all the time, its just the few times that i notice the traffic getting high.  the older servers i have are the dual nic, but all the newer servers ive got are gigabit adapters.  we just cant justify the cost yet of a 3750 gigabit switch.  i was hoping to pick up a little bit extra with the equipment that weve got.

well, i was able to get this in on a server thats not used as much, and the results werent good.  i had to remove because some clients werent able to connect to that server while others still were.

Suggest you  read up on Intel's document and understand the contents before you attempt this on a production network. You may also check the server manufacturer's site for more info.

well, thanks for all the info guys.  ill split the points.