It’s been… well a while since Avaya presented at #WFD7 and I find myself thinking once or twice a day about SPB networks and how someone (maybe even me) might build a campus network using this technology.
I know there are a few delegates out there who weren’t overly enthusiastic about Avaya’s very network-centric presentation and wanted to “Get to the Wi-Fi.” But one thing we sometimes forget is that there is no Wi-Fi without the underlying network for that data to run on. I honestly think the days of wired and wireless networks being built separately are coming to an end. And as “Network Engineers” we must all think about the underlying architecture of the whole network.
Avaya’s Fabric Attach solution is a very complete package from a datacenter to campus SDN solution. From one perspective, the idea of creating networks that span entire campuses is very attractive. It’s one of the things that traditional controller-based networks do in order to simplify IP address and VLAN management. However SPB achieves this without the concept of tunneling all traffic to a central point and is not simply a network management system taking care of all the config for us.
I really dig their ability to offer Layer 2 and Layer 3 service as close to the client as possible. No need to “trombone” data, when you can deal with as close to the edge as possible. Also the idea of abstracting the network away from the client so they only see their default gateway is attractive to me when looking at client security and guest networks, two places I’m always looking at to evaluate potential threats.
The only thing that gave me pause when reading up on SPB, was the choice of Layer 2 vs Layer 3. While the Layer 2 wireless engineer in me can sure see how routing everything with IS-IS at Layer 2 seems attractive, part of me really thinks that Layer 3 (IP) is where routing should be implemented. While SPB uses a MAC-in-MAC method, it is entirely possible to leverage a MAC-in-IP overlay to accomplish the same principal (VxLAN anyone?). But ultimately, does the underlay network really matter that much in an SDN overlay world? IS-IS is still routing packets/frames from one place to another. So all-in-all a real mute point. The biggest disadvantage to a L2 is a lack of L2 tools to diagnose underlay problems.
Overall, I like the simplicity of Avaya’s solution, one protocol to do all the heavy lifting, pick optimal paths and keep services functioning as close to the client as possible. All good messages, and great for networks supporting large-scale Wi-Fi deployments from a solid architectural network design. I personally think they have a great message and are definitely committed to IEEE/IETF standards around SPB.
If you are looking for more information on Shortest Path Bridging: Paul Unbehagen (@punbehagen) has a pretty awesome blog over at RandomBytes. Hopefully we can encourage him to keep posting good content.
One note: Paul Unbehagen was pretty adamant about getting rid of tunnels, and this confused me a little bit. I was under the impression that SPB and 802.1aq used a MAC-in-MAC encapsulation technique. I know that it can inter-operate with other switches using the VLAN tag. My impression was that Avaya was using the MAC-in-MAC method for some of their Service Provider offerings so maybe he can jump in and comment. Just remember, I’m not the SPB expert.
I've seen Paul's excellent SPB talk a couple of times, and the general idea as I understand it is that when the tunnels are sufficiently ubiquitous and trivial to set up, you can pretend they're not there.
More seriously, the general idea is that you create a pseudo-ethernet underlay. The pseudo part allows it to scale beyond traditional ethernet by forwarding only between network devices using ISIS to generate forwarding tables, allowing the underlay to be broadcast free, and therefore loop safe. This L2 routing table plus encapsulation is roughly equivalent to creating a full mesh of tunnels between all of your SPB devices, but it's all completely managed dynamically via ISIS, avoiding the usual pains of managing a sea of tunnels.
For those who remember it, the end result is fairly close to an ATM LANE setup, though thankfully without the dedicated LES/BUS requirement. There are some pretty good videos from Avaya at tech field day focusing on SPB.
Layer 2 troubleshooting can be accomplished with layer 2 traceroute and layer 2 tracetree.
Yes, you can do layer 2 traceroute and layer 2 tracetree, but these are not tools that most network engineers use on traditional networks. My opinion is that a underlay based around IP gives you traditional tools for underlay troubleshooting.