Ether Vs No Ether

alliasneo

Hi guys,

I've only just started studying for the CCNP Switch and I've been messing around on my home lab with Etherchannels. From what i've read I can understand that your not going to get double, triple the throughout etc when you start bundling links together. I have put together the following topology using two switches:



I'm using FA0/10, 11 & 12 between the two and have put them in to one Etherchannel making the cost 9. So basically I've got a 300mb link between the two that traffic can be shared over.

So I ran Microsoft Sharepoint and transferred a 20mb filed between my laptop and my PC and recorded the results using windows task manager:




I'm just interested in what you think to the results? I wasn't expecting a huge peak between the two but I did expect the transfer to be noticeably different after I tore down the etherchannel. I'm wondering if there is more load balancing I need to configure to make the throughout more noticable or if what I have done is pretty accurate?

I can post the configs up if neccessary but I've just create a group-channel and made them trunks....


thanks

DPG

The load balancing is determined by either the source or destination address (mac or IP). You are only going to see one link utilized for a flow from one host to another.

routergods

There are two issues here....

1. The default load balancing on etherchannel is by source-mac address. Since you're just transferring between one computer to a "server", the source mac isn't changing. Therefore, it's not going to load balance. Now if you added another laptop and ran a simultaneous transfer, you probably would see better results. You can change to other load balancing options like dest mac, xor combination of dest/source macs, ports, etc... but that won't help you if all you are doing is sending one stream from one pc to another.

2. Never use three ports for etherchannel because the split isn't even. I think it's 30%/20%/20% when you use three links.. plus the hashing algorithm throws a fit.

Zartanasaurus

Any individual flow is not going to get increased bandwidth any more than your car could drive faster on the highway because another lane is added (ignoring the effects of reduced congestion in your lane).

NetworkVeteran

Zartanasaurus wrote: »

Any individual flow is not going to get increased bandwidth any more than your car could drive faster on the highway because another lane is added (ignoring the effects of reduced congestion in your lane).

Excellent comparison!

sratakhin

Etherchannel won't help you when you only have two hosts. I tested bundled connections between two switches and the effect is only noticeable when you have at least 4 hosts transferring at the same time.

alliasneo

thanks very much for the responses everyone.

routergods - thanks for letting me know about the x3 ports. That makes perfect sense.

Zartanasaurus -thanks. This is very easy to understand. That's makes sense though, so if add more traffic that's where the other channels would be utilised thus making performance of the links shared and more overall bandwidth for everyone as opposed to using just one link.