11-17-2010 07:40 AM
There was an interesting debate going on yesterday in Tweetville over the history of "End-to-End" FCoE and the meaning of the title of an article by Mike Fratto of Network Computing, "Brocade First To Market With Native End-to-End FCoE". In this mornings followup article, Mike clarifies his claim. He provides a summary of how the Brocade VCS technology handles FCoE traffic in an Ethernet fabric, in essence, he says Brocade has the first end-to-end Ethernet fabric that carries FCoE traffic.
In fact, the Brocade "First" is the Ethernet fabric, something new that solves multiple problems with ETHERNET, not just FCoE. The tweeting got so loud, (I had to put ear plugs in to keep within OSHA guidelines ;-) ) that point was overlooked.
An Ethernet fabric is good for ... well, Ethernet, of course. Since FCoE requires Ethernet to transport Fibre Channel frames, you would expect it to flow across an Ethernet fabric, and it does. But, that requires the Ethernet fabric to support the Data Center Bridging (DCB) extensions so Ethernet is "lossless, and in fact VCS technology includes the necessary DCB extensions for lossless Ethernet:
Mike also points out some of the key features of an Ethernet fabric with VCS technology.
"Brocade's VCS uses multiple paths at the frame level, meaning the same flow can be distributed across multiple links, which Brocade does for FC as well."
That's automatic formation of inter-switch links AND automatic trunk creation. These are very cool features of VCS technology that ensure the highest bandwidth utilization of multiple equal cost paths between switches while eliminating manual configuration and mistakes. And, this is NOT restricted to FCoE traffic. These features are built into VCS technology so ALL Ethernet frames can take advantage.
Mike also makes another good point.
"Just to put a bow on this monster: two nodes using FCoE connected to a Nexus 2000 Fabric Extender, which is connected to a Nexus 5000, does not constitute end-to-end Ethernet FCoE because the Nexus 2000 Fabric Extender is just a bump in the wire and switching occurs on the Nexus 5000."
That's what we mean about being first to market with an Ethernet fabric. Just to refresh some of the key features of an Ethernet fabric:
11-23-2010 01:08 AM
You say the trunks are created automatically. Ethernet switches supported LACP for quite some time now and this protocol requires little to no configuration. Why is this is a big feature of VCS? And, most stacking products support automatic stacking link formation. Can you please explain how VCS is different from the industry standard switches in this respects?
11-26-2010 02:45 PM
Here is my 2 cents.
LACP is like the ISL (or automatic trunking) to a point.
I ould think the main differance is in that if take 4 or 8 interfaces and LACP them together you are creating a 4GB or 8GB (ok or 40GB or 80GB) logical single link. Now even though we now have this larger single link we must still follow the session rules with LACP.
This means any one traffic flow cannot use more then a single phyical link in the LACP. Now this is where the ISL come into their own. Any flow can and will go over any phyical link in the ISL.
Any yes setting up a LACP is simply and easy - ISL are automatic = this can be both good and bad in my mind.
The good is 'it just works'
The bad is when plugging the switches together a basic understanding of port group should be know as connecting port from different port group will end up in extra ISL being created.
Stacking vs VCS
Like the LACP story, the VCS expends on the idea of a stock - by not only knowning about the other switches in a stack and giving single point of management but by also knowning and tracking the devices that are plugged into the frabic. Think Virtual machines as an example - Vmotion (or Live mirgration) move the guest to another host - all good. Now with VCS we can track that move and apply the same switching requirments on the new host for thing like VLANs, port security etc.
Note - I do not work for Brocade - any errors provided for free - please correct me if I am wrong.
11-30-2010 08:12 PM
Thanks mschipp. That helps. I also see another blog article from Mr. Brook Reams explaining this in more detail. Thanks to both of you.
Just a minor clarification. Since stacking solutions have a single master, the master knows of all the devices connected to the stack and therefore can track those devices equally well. Also, I am not aware of any stacking solution that actually trunk multiple stacking links between the stacking units to provide more east-west band-witdh. The flexibility, as explained from Mr. Brook, and the performance of these trunks seem very interesting features.