The state of TRILL

I often get three questions about TRILL: Are the TRILL standards finalized? Has anyone implemented it? Is it useful?

Short answers: Yes, No, Maybe (although I remain unconvinced).

Are the TRILL standards finalized?

The TRILL working group has produces several RFCs and is still working on numerous draft documents. The core RFCs it has produced (RFC 6325 – RBridge Base Protocol, RFC 6327 – RBridge Adjacency and RFC 6439 – RBridge Appointed Forwarders) are all Proposed Standards, and together with RFC 6326 (TRILL use of IS-IS) give enough information that someone could start writing a TRILL-compliant implementation.

Some people think that the Proposed Standard status of these documents means the standards are not finalized. If you’re one of them, read the RFC 2026 (Internet Standards Process) first – the proposed standard status indicates a standard’s maturity level.

Every IETF standard starts as a proposed standard until there are at least two independent interoperable implementations, at which point its status could have been changed to draft standard (no longer – RFC 6410 removed the draft standard maturity level). Only when a technology reaches a “high degree of technical maturity” and “provides significant benefit to the Internet community”, the corresponding RFC becomes an Internet standard.

Just to give you an example: RFC 4862 – IPv6 Stateless Autoconfiguration (SLAAC) is still a draft standard, although there just might be more than a few interoperable implementations, and it happens to be occasionally used in a few networks.

Summary: the core TRILL standards are finalized, but of course the TRILL working group is still developing numerous add-ons.

Has anyone implemented TRILL?

Do you mean the likes of HP and Force10 that have been talking about TRILL for years? They’re still living in the TRILL promiseland.

I might have missed a vendor launching a TRILL-like solution, but so far I’m aware of only two vendors with something remotely similar to TRILL: Cisco’s FabricPath (which, like TRILL, uses IS-IS, but has its own encapsulation format) and Brocade’s VCS Fabric (which uses TRILL packet format and a proprietary version of FSPF as the routing protocol).

Summary: Nobody has implemented the core TRILL standards in a way that would have a remote chance of being interoperable with anyone else.

Is TRILL useful?

You haven’t been reading my blog lately, have you? Why don’t you start with Does TRILL make sense at all?

The original RBridges idea Radia Perlman had (assuming I understood it correctly), which would effectively turn an IP subnet into a CLNP-like IS-IS area, would be interesting (more so if the RBridges would have enough IP awareness to act as ARP proxies and reduce the broadcasting). Unfortunately her simple idea deteriorated in the TRILL working group as soon as networking vendors started slapping all sorts of VLAN-related concoctions on top of it (watch her reaction in this GoogleTalk video when someone asks a VLAN-related question).

TRILL would have been useful five years ago. As things stand right now, the primary use case (large VLAN-based virtual networks within a data center) will no longer be relevant well before we’ll see two interoperable TRILL implementations.

8 comments:

  1. How about to use TRILL in IXP?

    ReplyDelete
    Replies
    1. You could, but VPLS gives you way more options (including MPLS TE and FRR).

      Go through the comments to my recent IXP post, they include some links to great presentations describing how large IXPs work.

      Delete
  2. Still unclear how a Vswitch removes the need for physical switch inter-connectivity loop avoidance.

    ReplyDelete
    Replies
    1. All hypervisor vendors are going in the direction of running L2 networks on top of IP. Hypervisor servers become simple IP hosts. No loop avoidance, no bridging, just IP, pure & simple.

      Delete
    2. Thanks for answering.

      Right but the hypervisor server still needs to be connected to a physical switch right? That's connected to a bunch of other physical switches to eventually connect to the WAN/Internet. Those servers have MAC addresses and are connected by ethernet. I just do not get it.

      Delete
    3. Elementary - it's called IP routing. You configure switch uplinks as L3 interfaces and inter-switch links as IP subnets.

      How do you think Internet works? It's not one huge bridging domain.

      Delete
  3. So we have 2014 and I´m still waiting for "cheap" TRILL/SPB switches. How long will it take till I can buy noname switches in Walmart?

    Btw, does anyone TRILL test units, i.e. boxes with ~4 ports (may it be 10Mbit/s) just for playing with...

    ReplyDelete
    Replies
    1. I wouldn't hold my breath waiting for low-cost TRILL boxes ... but then HP has TRILL on the 5900 switches.

      Delete

You don't have to log in to post a comment, but please do provide your real name/URL. Anonymous comments might get deleted.

Ivan Pepelnjak, CCIE#1354, is the chief technology advisor for NIL Data Communications. He has been designing and implementing large-scale data communications networks as well as teaching and writing books about advanced technologies since 1990. See his full profile, contact him or follow @ioshints on Twitter.