Don’t lie about proprietary protocols
A few months ago Brocade launched its own version of Data Center Fabric (VCS) and the VDX series of switches claiming that:
The Ethernet Fabric is an advanced multi-path network utilizing an emerging standard called Transparent Interconnection of Lots of Links (TRILL).
Those familiar with TRILL were immediately suspicious as some of the Brocade’s materials mentioned TRILL in the same sentence as FSPF, but we could not go beyond speculations. The Brocade’s Network OS Administrator’s Guide (Supporting Network OS v2.0, December 2010) shows a clear picture.
The TRILL section of the Fabric chapter (page 89) claims that:
The Brocade Ethernet fabric uses Transparent Interconnection of Lots of Links (TRILL) protocol, designed for the sole purpose of scaling Ethernet networks by allowing a set of devices, called routing bridges (RBridges), to connect with each other.
However, the very next paragraph explains that:
Link state routing in VCS-based TRILL networks is performed using Brocade’s proven Fabric Shortest Path First (FSPF) protocol.
I understand that Brocade has way more FSPF than IS-IS experience, but the RBridges: Base Protocol Specification is very clear. In section 4.2 it states that “TRILL uses an extension of IS-IS [ISO10589] [RFC1195] as its routing protocol.”
There are numerous other details where the concepts of VCS fabric deviate from TRILL. For example:
- RBridge ID is assigned through Fiber Channel Request for Domain ID and Domain ID Assignment messages. RBridges: Base Protocol Specification uses a completely different process to select RBridge ID (Section 4.2.1) and RBridge Nickname (Section 3.7.3).
- Switches use Brocade Link Discovery Process (BLDP) to discover other switches in the fabric. TRILL uses IS-IS hello messages.
VCS fabric does use TRILL framing, as explained in the VCS Ethernet Fabric and FCoE Traffic article by Brook Reams, and according to the same author Brocade plans to replace FSPF with IS-IS in the future. Brocade’s engineers are very clear about what VCS fabric does and does not do, but the marketing claim that VCS uses TRILL is a clear misrepresentation that casts unnecessary doubts on the new products and raises confusion where none should have existed. Cisco was at least honest enough to admit that FabricPath is not TRILL (it does use TRILL concepts, including IS-IS, but not TRILL framing).
It’s always debatable whether it makes sense to use proprietary solutions or not (and whether they're harmful), but the networking world is full of them and we all have to live with them. However, it’s crucial to understand which parts of a solution (or which protocols) are proprietary and which are standard and interoperable with other vendors. It’s totally inappropriate that one has to dig pretty deep into Brocade’s web site to discover what the true story about TRILL support is. We expected more from an established networking vendor.
Update 2011-03-05: inserted new facts described in Brook Reams’ blog posts and reworded the conclusions.
Ok that me laugh, because some manufacturers that make "carrier grade" equipment can't get the standards right so that their own equipment works properly, let alone with another manufacturers equipment. When you phone them and they say "no, no, no don't use MSTP!" you realise that even they know their implementation is badly broken.
Why half implement a standard? Don't they realize it's not standards compliant if they only half implement it? I'll take a working proprietary protocol over a broken standards implementation any day!
What this also says to me is that Brocade is taking the code they have (which presumably at least works) having seen the difficulty of building new protocols they historically may not have in their code base. Could they be planning on shifting TRILL to IS-IS later? Is this a time-to-market move? (Tempted to say, kluge -- one could argue one SPF is as good as another I suppose, not sure I'd buy that argument.)
How is using a TRILL frame in the data plane lying? How is clarifying the use of FSPF in the control plane a lie? How is talking about a roadmap for future IS-IS in the control plane a lie? I don't get it. It sounds like Brocade is being very explicit and truthful.
First of all, thanks for your feedback!
Saying "we use TRILL framing in the data plane" is perfectly OK. I must have missed this very precise definition in your whitepapers; if that's the case, I apologize - I would also appreciate a link to the document describing exactly which parts of your implementation are complaint with TRILL.
Saying "Brocade fabric uses TRILL protocol" is not OK. TRILL has numerous components and in my humble opinion you cannot claim to "use TRILL" if you use just the framing format. It's like claiming you use OSI protocols just because your link-layer encapsulation is LAPB (and you might run IP with OSPF on top of that).
I cannot comment on whether "clarifying the use of FSPF in the control plane" is or is not a lie; I assume it's true.
If you would have said "we're already using TRILL framing format, so we will be ready to deploy TRILL when the IS-IS-related drafts are ratified and will require only software upgrade; in the meantime we decided to use FSPF" I would have applauded your honesty. Your marketing people have unfortunately decided to claim something completely different.
Indeed, "utilize" can be misconstrued, and we have stated that VCS Technology "utilizes" the TRILL standard. In particular, we utilize TRILL frames for the data path but currently use FSPF in the control plane, along with a number of other unique capabilities in our VCS Technology.
To your comment, "If you would have said "we're already using TRILL framing format so will be ready to deply TRILL when the IS-IS related drafts are rarified and will require only software upgrade; in the meantime we decided to use FSPF".
We did say that and continue to say that and continue to be as transparent and clear about how we utilize TRILL as we can.
For example, please read these publicly available postings.
And, in an attempt to clarify some ignorance that came up regarding VCS when Forrester's Andre Kindenss blogged about Juniper's QFabric announcement:
Ivan, I continue to appreciate your authoritative blogs. Keep up the good work my friend.
I can make a minor point of clarification to your comment "Could they be planning on shifting TRILL to IS-IS later?" and to Ivan's comment "I understand that Brocade has years of FSPF experience and zero IS-IS experience ..."
Brocade supports IS-IS today in our IP networking products and does understand this protocol. Please visit http://www.brocade.com/products/all/routers/index.page for more details.
What's needed is the ratification of the modifications to IS-IS for layer 2 support required to use it with the TRILL standard. Please see my comment above for more on that if you are interested.
Have a great weekend.
Thanks for all the links. They also clarified a few other points I was wondering about. I reworded the conclusion. It's obvious from your blog posts that you're trying hard to be perfectly transparent and clear (and I guess that - as is often the case - the marketing claims don't exactly help your efforts).
I have been reading various lovely posts on your blog about lovingly lovely lovely TRILL. So I got this impression that TRILL is all over & everyone seems to be talking about it (especially yourself & Greg). So if I am right, TRILL is just a work in progress as of now & the standard is not finalized yet & there isnt any vendor right now who is actually using TRILL already?
Can you please confirm that, Thanks