Solved: value of the STP's root bridge?

ocicat002 · ‎08-13-2008

Although I have found multiple sources which describes how the root bridge in a redundant switch topology can be determined, why does it matter? If I only have three switches all connected together, what does it matter which one is declared the root?

Thanks for any insight shared.

Joseph W. Doherty · ‎08-13-2008

For three switches connected in a triangle, one link of the three inconnecting links will block. You will then have one switch as the transit for the two others. Assuming one switch offers more performance than the other two, setting it to the root will then make it the transit switch. Or, assuming all switches offer the same performance, but two of the interconnecting links provide more bandwidth than third, you want the lessor bandwidth link to block. This too might be accomplished by making the switch that terminates both the higher bandwidth links the root switch.

View solution in original post

Istvan_Rabai · ‎08-26-2008

To add to Joseph's post:

The root bridge switch will have to bear all the traffic between the branches of the spanning tree.

So from design perspective it is better to choose the switch for the root bridge role which has the highest throughput and highest speed links.

Cheers:

Istvan

View solution in original post

paul.matthews · ‎09-04-2008

And another thing...

If you don't care where the root is and let it sort itself, if one switch (normally the root) has failed, one of the others will take over as root. When the normal root returns, it will assert itself as root, which will be more disruptive.

That might not sound much, but I had an interesting one for an implementation recently. On any VLAN three switches were involved - two core switches and an access switch. The customer was keen that most fault situations would not cause a loss of a phone call, and where possible all core-access switch links should be load balancing. So, we had a L3 link between the two core switches. The roots were forced to the core switches initially and we had GLBP tuned right down.

We noticed that on restoration of a failed link between a core and an access switch, where the link was to the primary root, even using RSTP there were a few packets dropped, as the root bridge moved from one end to the other. Our solution? we made access switches the root switches.

View solution in original post

Joseph W. Doherty · ‎08-13-2008

For three switches connected in a triangle, one link of the three inconnecting links will block. You will then have one switch as the transit for the two others. Assuming one switch offers more performance than the other two, setting it to the root will then make it the transit switch. Or, assuming all switches offer the same performance, but two of the interconnecting links provide more bandwidth than third, you want the lessor bandwidth link to block. This too might be accomplished by making the switch that terminates both the higher bandwidth links the root switch.

Istvan_Rabai · ‎08-26-2008

To add to Joseph's post:

The root bridge switch will have to bear all the traffic between the branches of the spanning tree.

So from design perspective it is better to choose the switch for the root bridge role which has the highest throughput and highest speed links.

Cheers:

Istvan

paul.matthews · ‎09-04-2008

And another thing...

If you don't care where the root is and let it sort itself, if one switch (normally the root) has failed, one of the others will take over as root. When the normal root returns, it will assert itself as root, which will be more disruptive.

That might not sound much, but I had an interesting one for an implementation recently. On any VLAN three switches were involved - two core switches and an access switch. The customer was keen that most fault situations would not cause a loss of a phone call, and where possible all core-access switch links should be load balancing. So, we had a L3 link between the two core switches. The roots were forced to the core switches initially and we had GLBP tuned right down.

We noticed that on restoration of a failed link between a core and an access switch, where the link was to the primary root, even using RSTP there were a few packets dropped, as the root bridge moved from one end to the other. Our solution? we made access switches the root switches.