Link aggregation with 4 copper cables between 2 PowerConnect 5324

Question

Hello,

I try to create a 4GB trunk between 2 Powerconnect 5324 using 4 copper cables connected on the ports 17 to 20 on both switches.

Here is my configuration for SWITCH_1 :
interface ethernet g17
channel-group 1 mode on
no negotiation
duplex full
speed 1000
exit
interface ethernet g18
channel-group 1 mode on
no negotiation
duplex full
speed 1000
exit
interface ethernet g19
channel-group 1 mode on
no negotiation
duplex full
speed 1000
exit
interface ethernet g20
channel-group 1 mode on
no negotiation
duplex full
speed 1000
exit
interface port-channel 1
description Switchs:1-2__Ports:17-18-19-20__Speed:4GB
exit
interface vlan 1
ip address 10.1.200.253 255.0.0.0
exit

Here is my configuration for SWITCH_2 :
interface ethernet g17
channel-group 1 mode on
no negotiation
duplex full
speed 1000
exit
interface ethernet g18
channel-group 1 mode on
no negotiation
duplex full
speed 1000
exit
interface ethernet g19
channel-group 1 mode on
no negotiation
duplex full
speed 1000
exit
interface ethernet g20
channel-group 1 mode on
no negotiation
duplex full
speed 1000
exit
interface port-channel 1
description Switchs:1-2__Ports:17-18-19-20__Speed:4GB
exit
interface vlan 1
ip address 10.1.200.252 255.0.0.0
exit

While connected via telnet to one of the switches, if I check the configuraiton of the interfaces with the CLI commands, I can see that my trunk is not 4GB but 1GB :

switch1# show interface configuration
Flow Admin Back Mdix
Port Type Duplex Speed Neg control State Pressure Mode
........ ............ ...... ..... ........ ....... ..... ........ ....
(...)
g17 1G-Copper Full 1000 Disabled Off Up Disabled Auto
g18 1G-Copper Full 1000 Disabled Off Up Disabled Auto
g19 1G-Copper Full 1000 Disabled Off Up Disabled Auto
g20 1G-Copper Full 1000 Disabled Off Up Disabled Auto
(...)
Flow Admin Back
Ch Type Speed Neg control State Pressure
........ ....... ..... ........ ....... ..... ........
ch1 1G 1000 Enabled Off Up Disabled
(...)

Any clue? Did I forget to configure anything? What is wrong with my configs?
I would really appreciate if anyone could help me.

DELL-Cuong N. · Answer

There is nothing that would show you 4GB speed.  LAG is a logical bonding of multiple ports to achieve an overall traffic flow between two switches that would improve the overall performance over the trunk port but you cannot send more then 1GB over any physical Gig port.  Also please note that for any given data stream between two systems (a server and a given host) you will not get any more then the speed of the physical port.  The idea though is that over the trunk between the two switches if you have lots of different sources and different destinations you are not trying to get all those traffic over the same single 1GB trunk.  The key to note here is that from any given source/destination even if your traffic exceeds the single port physical BW the excess does not 'flow' over to the other ports in the LAG.   Please review this paper which provide a very good explanation of how LAG works.   http://www.dell.com/downloads/global/power/ps2q05-20040286-Holmes-OE.pdf   Cuong.

parioca · Answer

Thank you for your answer, Cuong.

The reason why I want this 4 ports LAG between my 2 switches is I need a bandwitdth larger than 1GB between them, even though it is a logical one. I have well understood that I will never be able to send more than 1GB on one port because it is the physical limit. I just need to allow a maximum of servers connected to the first switch to 'discuss' with a maximum of servers connected to the second one, in the mean time, and with no bottlenecks if possible.

In that perspective, could you please tell me if I configured my switches correctly? Or maybe there would be a better practice? And is there a way to check that my LAG is 'large' enough ?

DELL-Cuong N. · Answer

I don't see anything wrong with the way you configured it. It is good that if you disable auto-negotiation on the port that you also make sure to configure both sides of the link to the same speed/duplex otherwise it causes some weird problems but anyway since you made sure that the links are configured the same way it is fine.

Once you setup the LAG, the only way I can think of to see what kind of throughput you are getting is to run some tests and see. You expect that for a random set of source/destination (again please review the link to the paper I referenced earlier), you should generally see a statistically good distribution of those streams of packets over the four links on your LAG. You may see all the streams on the single link and you may see one or more link(s) not used even as the others are overloaded. There is no guaranteed of a perfect distribution because of the way LAG works (again please read the paper). But in general, for a random mix of source/destination you should get statistically good distribution of the traffic streams over the four links.

Cuong.

Networking General

Link aggregation with 4 copper cables between 2 PowerConnect 5324

Was this post helpful?