LAG from one server to two switches?

[Michael BUTOW]

Hello,

Is it possible to set up a link aggregation group from one server to two switches?

I need redundancy in case of a switch failure, but NetRAIN failover times are too large for my purposes - that is why I am interested in alternatives.

Best regards,
Michael

[Mark Poeschl_2]

No it's not. To make LAG work you need a pair of ports on a single device that can handle one of the link aggregation protocols like LACP or PAGP.

[Al Licause]

Mark is correct....but it is a limitation imposed by the switch vendors. So if it is a necessity, you might want to poll different switch vendors to see if anyone has yet implemented LAG across switches.

[Michael BUTOW]

Thanks for the information. We have already a fixed hardware configuration (cabinets) with two HP ProCurve switches in each cabinet(2x2524 or 2x2824).
I could not find anything about this topic in the documentation of these switches (the presented drawings always show only one switch). Does anyone perhaps have experience with LAG across two of these (of the same device type though).

[Al Licause]

Since this forum is specific to Tru64unix network issues, you might want to post to the
networking switches and stuff forum. They would be better suited to answer switch questions.

Good luck.

[Mark Poeschl_2]

But think about it, for this to work the two switches in question would need either a) some out-of-band interconnect to keep track of what's happening on the aggregated link, or b) to use a substantial amount of their in-band capacity to communicate with each other about the aggregated link. Option b is really ugly since it clogs your network with aggregated link status traffic and option a is gonna get pricey really fast. It's not at all surprising nobody's doing this.

[Michael BUTOW]

Okay, I will put this question about the ProCurve switches into a more appropriate forum.

Thanks for the interesting comments and advice - I guess I may have to just experiment with it a bit.