Vlan routing 2910s and 1810s into a 2920 stack

[Galations220]

I have 6 vlans defined on the 2920 stack.

what are my options for connecting the 2910 and 1810 to them .

Each 2910 or 1810 only has 1 vlan, maybe 2 if I can reasonably work out a management vlan.

In the future, I could also see the need for a voice vlan that would span all the switches.

Can I set the 2910s up for vlan tagging and then switch to routing at the 2920? 

I have a support ticket with HP outstanding which says if I trunk(LACP 2 ports) to the 1810 or 2910 to the 2920, I need use vlan routing.  Is this true, or could I just leave the trunk untagged for the primary vlan and layer 3 route at the 2920? 

or best to go vlan routing using layer 2 all the way?

is layer2 vlan routing much faster than layer3 routing?

[Vince-Whirlwind]

Switching (Layer2) should always be preferred over Routing (Layer3) wherever possible.

If you have three switches, a 2920, 2910, and 1910, and you require routing on your network, then you should pick one device to be the Layer3 switch, and make the other 2 devices Layer2 switches.

You should pick the device with the best performance and features to be the Layer3 switch.

In this case, that is the 2920. This is your "Core" switch.

This means your other two devices should be Layer2 only.

Each device should have an IP address in the Management VLAN. The 2910 and 1910 should have no other IP addresses configured on them.

For every subnet (VLAN) required on each switch, you need to extend the VLAN from the 2920 to the other switch.

To extend a VLAN, you need to

On the 2920:

 - create the VLAN on the "Core" switch. Assign it the router (default GW) address for that VLAN's subnet.

 - add the VLAN as a tagged VLAN to the downlink port to the 2910

On the 2910:

 - create the VLAN

 - add the VLAN as a tagged VLAN to the uplink port to the 2920

 - for each host that is patched to the 2910 that requires that VLAN, add the VLAN as an untagged VLAN to the switchport it is patched to.

[Vince-Whirlwind]

LACP is a completely seperate issue. Don't worry about it until you have your VLANs configured and have confirmed connectivity is established.

[Galations220]

I think my configuration is setup how you described. 

Ports 11 are vlan 11 on the 2920 stack

Ports 12  are vlan 12 on the 2920 stack

Ports 11 and 12 are lacp trunked into the vlan switches.   Traffic is flowing, but here is the brain teaser...  RDP from vlan 11 to 12 is really slow.  However the reverse is not true, 12 to 11 is really fast.  The difference is vlan 11 has an extra 8 port cisco desktop switch in the mix.  If this switch is removed, RDP is fast in both directions. 

Another piece of the puzzle , RDP from vlan 11 to vlan 1 is fast, even with the cisco desktop switch in play, its the extra hop that really slows it down for some reason.  Specifically, the RDP logon window responds almost instantly, but the authentication can take up to 30 seconds. 

Windows is reporting fire transfers to be wire speed  or close to wire speed (120 MB/s) from any vlan to another, so it is just the RDP protocol that is slow.

IP routing is enabled on the 2920 switch

                                            tagged (vlan 11)

2920 (vlan 1) stack  -------------------- > 2910 switch (vlan 11)

          |

          |

          |  tagged for (vlan 12)

          |

          |

1810 switch (vlan 12)

I think I could turn off IP routing and go strictly vlan routing, but I will need to wait for the weekend to try that.  However it sounds like you think this configuration should be workable as is?   And for the most part we dont have any problems, but we really use the RDP protocol a great deal = (

[Galations220]

One more note. After RDP authenticates, it is fine, there is no delay.  It is just the authentication process that can take half a minute or more.

[Vince-Whirlwind]

Doesn't much sound like a network issue.

Sounds more like the subnet you have on that VLAN isn't in AD properly

What about other authenticated services from that subnet - eg web-proxy?