Blade server switch connection to Catalyst switch - design considerations

[fawad.alam]

Does anyone have experience configuring the IBM blade server switch (running Cisco IOS ) and configuring it to work with any other Cisco uplink switch? I am having some issue with configuration and need to understand how the port mapping and failover works. Any assistance will be appreciated!!!

[glen.grant]

We generally take 2 of the uplink ports and etherchannel and trunk them to one switch and the other 2 are etherchanneled and trunked to a different switch . The blades themselves are trunked to allow all vlans across all 14 blades then you don't have to mess with them again . Each port is mapped to a blade , g0/1 switchport is tied to blade #1 etc. Just think of each port as a normal switchport that is connected internal to a server instead of the normal external switchports to a standalone server. Those switches are basically a 2950 series switch in card form.

[fawad.alam]

Hi Glen, thanks for your input. For the connectivity option you have mentioned here, does it have both uplink ports in the forwording mode? Is there any chance of looping?

Would you have sample configs for both uplink and downlink ports? I would be interested to know about the aggregation switch configuration, especially the trunk and spanning tree details.

Thanks much !!!!

[Konstantin Dunaev]

just consider them as the normal cisco switches and each server is connected to both switches.

you can connect each blade switch to your both uplink switch(2+2)

or you can connect each blade switch to only one uplink switch and interconnetc the blade switches

or you can connect all uplinks ports of blade switch to a single uplink switch, configure the "track interface" option on bladeswitch to disable the ESX's ports if your uplink port is down. Or configure the ESX servers to understand somehow the STP messages (I think it should be possible to do in the new version) and use the interfaces accordingly.

One thing to mention - you should have a look how the ESX servers are sending the traffic, if ESX uses load-balancing mode you can get a lot of "unknown unicast" traffic on your uplink switch.