Comware

Hi,

In about 10 days we'll receive a couple more of the ProCurve 5372XL (5308XL) devices. I'll will need to set them up with trunking but am wondering:

Which of the solutions mentioned in the subject is preferred between 2 5372's?

What is the more technical differences between these technologies? I know one is from Cisco blah blah blah, but in which way do they operate differently?

Thank you very much in advance...
Rasmus

Hi Rasmus
i think you mean trunk as portaggregation, then
trunk is manual form a trunk
LACP is the autmatic standardprotocol (802.3ad)
FEC is the cisco portaggregation protocol (cisco fastetherchannel).

if you connects 5308xl's to 5308xl's is switchmeshing a good idea.
meshing is described in the manual and easy to use. it is a loadbalancing technology in a mashing switch environment.

good luck
carsten

Hi - and thanks for your answer.

We will not be using meshing since we will have routing enabled on all switches. Meshing is not supported with routing enabled.

So when I create the trunks between the switches, should I use the Trunk, FEC or LACP option?

I know that FEC does not balance the traffic on the trunks, but instead the connections. How does Trunk and LACP work? Do they balance all traffic, or distribute the single connections to the least heavy link?

Thanks in advance,
Rasmus

Hi Rasmus
Trunk, FEC and LACP use the same mechanism.
Outbound-Trafficdistribution based of the Source/Destination -mac-address -pair.
Use LACP on the 5308-Connectionpoints and disable LACP on all other.
The Management& Configuration-Guide described it in chap 9.

Carsten

I was under the impression that the LACP algorithm is better than the FEC algorithm in distributing load. SA/DA distribution is used by FEC.

The HP documentation states that all three methods of trunking (Trunk, FEC and LACP) use the SA/DA method for distribution.

I haven't done extensive testing in a switch to switch configuration, but in a switch to server configuration load balancing does not appear to work correctly on the Procurve line.

Does anyone know where we can find the algorithms that determine how traffic is distributed amongst the members of the trunk? For both FEC and LACP.

Hi,
I'm trying to setup a trunk on a new Proliant DL580 with a card with 2 Gig Copper Port on a 4108Gl switch.

When i configure the trunk on both side, if i user "Trunk" method then all other Gigabit copper card (some DL380) cannot be reach (even if they are plugged on another switch), all other (Fibre card and 10/100 Card can ben reach correctly) i even tried to plug a 1000G Copper on a 10/100 Port but i can't reach it.

It is the same thing if i use FEC trunking but when i use LACP it is working fine.

is there anyone that can explain me why? My switch is updated to G_07_53 and all my drivers are up to date.

If you're connecting 2 5300xl's together via a trunk, I would recommend using the 'Trunk' option. This is the normal, basic trunking mechanism used by ProCurve switches and it does not send any protocol traffic across the trunk. LACP/FEC both send protocol traffic across the trunk to negotiate.

Hi all,
but after all post I don't undersatand if the trunk protocol can aggregate a channel e if the trun can balanced the network traffic.
Can explain me?
Thanks
Ciao

lucas-one,

if you trunk (HP-wise), or aggregate say 4 individual 100 Mbps links into a LACP or FEC trunk between 2 HP switches, you'll have a theoretical 800 Mbps FDx (400 Mbps down-trunk + 400 Mbps up-trunk, so to speak).

LACP is IEEE-standards-based Link Aggregation Control Protocol, whereas FEC is Cisco-proprietary Fast EtherChannel.

From previous posts on this subject, it seems LACP sends everything down one physical link in the trunk if it has to handle say 95 Mbps of down-trunk traffic, and uses 2 phys. links if it has to handle say 175 Mbps - and so on: it uses more phys. links until the max capacity of 400 Mbps oneway-trunk. It does not seem to balance the 95 Mbps over all the 4 phys. links. I guess you could look at the internal workings/specs of the standard, if these are public domain, and posted by IEEE.

Seems that FEC does this balancing regardless of the data flow it has to handle, and is more capable than LACP in this respect (and others, I would say), however since it is proprietary, and NOT recommended anymore by manufacturers (Cisco included) for interop reasons, it could be harder to find the specs to have a look on them - to clear the air on what this protocol actually does.

Enough?

Well i wonder about the same... HP implementation of trunking technique seems to be strange.

If you choose Trunk: no loadbalacing
If you choose LACP: no loadbalacing
If you choose FEC: no loadbalacing

when i use FEC between 2 Cisco switches i get a load balance.

So it seems there is no possibility for loadbalacing on HP trunks?