switch cbs 250 - lag

[tecnicOIT]

Two switches connected in a lag with the two fiber sfp ports.

If I also connect them with ethernet cables using 1 or 2 or 3 ports and then add them to the lag of the two fiber abobe, they still work but do you think that this improves the bandwith between the switches?

thanks

[balaji.bandi · Accepted solution]

basic LAG requirement, make sure both the side same speed (and member of LAG also same SPEED)

does not matter what media, i have mixed RJ45 and Fibre ports (same 1GB speed works as expected).

i suggest  always use even ports like 2 or 4  or 6 (to load balance as expected)  rather 3 or 5 or 7 ?

[M02@rt37 · Accepted solution]

Hello @tecnicOIT,

Creating a LAG with switches having the same speed and ensuring that the members of the LAG also have the same speed is indeed a fundamental requirement for LAG to function correctly. This ensures that all the links in the LAG can handle traffic at the same speed without causing issues. In many LAG configurations, load balancing is done based on specific algorithms that take into account the source and destination MAC addresses, IP addresses, or other factors. Using even ports can help ensure that the load balancing algorithm distributes traffic evenly across the members of the LAG. However, the exact behavior may depend on the specific switches and the load balancing algorithm they use. Some switches also allow you to configure how the load balancing is done, so you can choose the method that works best for your network's traffic patterns.

In general,:

-Use ports with the same speed.

-Use an even number of ports for load balancing purposes.

-Configure the switches to use an appropriate load balancing algorithm.

[Joseph W. Doherty · Accepted solution]

". . . but do you think that this improves the bandwith between the switches?"

It increases aggregate bandwidth, but individual flows are limited to bandwidth on one link, LAG (generally) does not account for link loading when directing flows (so for two links, you might realize about a 50% effective bandwidth boost) and flow distribution much depends on attributes being used for directing flows to links (best case, an average [over time] utilization spread over all the links; worst case, all flows get directed to the same link).