MPLS + P4 = Pensando
Leading up to Cloud Field Day 7, I was exceptionally excited about hearing from Pensando. Partly because I’m a bit of a neophile and they just came out of stealth. Also because what they are working on is really cool; super high geek-out factor – P4 in action baby! And not least of which, because you really just have to expect a lot from this team – MPLS is back at it!
So… Was I disappointed? Is the tech really as cool as it sounds? Can MPLS drive adoption without Cisco putting fuel in the tank? What do these questions even mean?
Read on to find out!
MPLS – Cisco = Sales?
I have to start by explaining something that probably is not clear to everyone. I’m not talking about Multi-Protocol Label Switching (MPLS) in this post. Rather, I’m talking about a group of four super-sharp (and super successful) folks: Mario Mazzola, Prem Jain, Luca Cafiero and Soni Jiandani (MPLS). With that out of the way, let’s begin.
For many start-ups most of us are totally blind to the capabilities of the leadership team. In some cases they are repeat entrepreneurs so you know they have one or more wins under their belt, and you’ve seen how they work. Pensando is a little different, they are multi-repeat founders, but with a twist.
Yes, they founded eight startups that generated billions of dollars. But one could argue that they did it with a bit of a safety net. While they did sell one business (Crescendo Communications) to Cisco more traditionally (back in 1993), the rest of their ventures were “spin-ins.”
MPLS + Chambers = Spin-in!
So, if a spin-out is when one company cuts a team lose to start a new company, what’s a spin-in? It’s similar but different. Essentially the MPLS team left Cisco multiple times, each time funded by Cisco to go start a new company with the express purpose of selling that company back into Cisco. They pioneered this approach to innovation with former Cisco Chairman and CEO John Chambers.
The advantage to Cisco was agile, start-up style innovation that could be targeted, if not outright shaped, by the networking behemoth. A clever solution to the innovators dilemma to be sure.
The advantage to MPLS was an (almost) guaranteed exit strategy, and a huge sales and marketing machine with a massive client list and gargantuan channel of value-added resellers (VARs). That’s got to make pushing new technology through the trough of disillusionment a heck of a lot easier.
The question we have to ask about Pensando then, is: Can they do it? Can this brilliant team sell a paradigm shifting technology to the world without Cisco’s resources?
One thing may tip the balance in their favor. While they don’t have Cisco, they do have the guy who was instrumental in building Cisco from the inside out for 26 years. If anyone knows how to sell new tech while scaling a company, it’s got to be John Chambers, right? Having him on the board is a very positive sign. And I’m sure no one involved would have it any other way.
Which is good, because this tech IS cool…
P4 + Hardware = Software-defined?
Hold up. What the heck is P4?
P4 is “a programming language for the data plan in networking devices such as routers, switches, and Network Interface Cards (NICs).” As Bruno Rijsman explains in his excellent Getting Started with P4 blog post.
Let me expound, in case that sentence didn’t blow your fuc*ing mind.
ASIC + Software = P4!
In the beginning there were ASICs. And ASICs were good. But, as with everything, their strength was their weakness.
The thing that makes an application-specific integrated circuit (ASIC) interesting is the application-specific bit. It means that particular chip was designed and manufactured for a specific application; tailored to a single use. This makes them very efficient at that task. It also makes them ill suited for other, even sometimes closely related, tasks. Like many ASICs that are awesome at VLAN encapsulation can’t do VxLAN routing at all, etc.
Even worse, designing a customized integrated circuit takes significant time, and spinning up a manufacturing plant to make that specific chip takes significant money. So the speed of innovation, and the freedom to innovate are both dampened. New features take a long time to come out, and people don’t “play” with any features that don’t have obvious mass market appeal.
Enter P4, and a whole new paradigm.
What if instead of burning fundamental network capabilities into silicon you could just write them in code? How would that impact the speed and breadth of network innovation? If you answered: “Wow!” we might be on the same page now. P4 is the programming language created for exactly that. P4’s four “Ps” are Programming Protocol-independent Packet Processors.
Hardware + Software = Pensando!
Now we can get to the next interesting question: Why is a company that claims to be all about “cloud” and “software-defined” selling hardware?
Justin is not wrong. And to be fair, we should look at Pensando’s whole pitch:
A software-defined, edge-accelerated, always-secure and visible, centrally managed platform that can run in any environment from the #1 innovation team in the world.
We spoke about the #1 innovation team above. And we just touched on the software-defined bit. But what about this edge-accelerated thing?
That, my friends, is in fact a thing; a hardware thing; a PCIe card (their DSC) to be specific.
And that card, combined with P4 and Pensando’s Policy and Services Manager (PSM), can do some amazing things.
To get a bit more specific; in one of their solution breifs they offer up some use cases:
- Distributed stateful firewall at the server edge
- Routing, Segment Routing, MPLS, BGP
- SDN and Virtual Networking with underlay/overlay encapsulations (VXLAN, etc.)
- Deep network and security visibility and telemetry
- East-West encryption within the data center
- Network load balancing, including TCP/TLS termination
- Storage virtualization functions including NVMe-oF
MPLS + P4 = Magic?
Which brings us to our final question; who’s going to buy it?
One market they’re going after is what I’ll call cloud-competitors. These are all of the IaaS/PaaS/BMaaS/FaaS/XaaS providers that aren’t one of the top 5-10 or so “public” clouds. And it makes a ton of sense. The Pensando system would be perfect for a lot of the things they want and need to do. I know it. At Myriad360 we call them cloud builders and we work with a bunch of them.
Of course, a complete architectural shift is a huge ask for a company already supporting (tens/hundreds of) thousands of customers. And most of these companies are building their infrastructure tech in-house. It’s an uphill battle for sure, but I don’t know of any better way to get such power and flexibility so deep into your infrastructure. Can we talk about end-to-end?
Another possible target is enterprises or institutions with high-performance compute (HPC) environments (and other specialized needs). Or even just massive private clouds (or even better, a set of giant old data centers that need to become a massive private cloud).
And I’m willing to guess that this is a product that might not have found its biggest use case yet. Pensando may end up providing the hardware and software for a new market. Maybe one created using their technology. Probably something with “edge” in the name of it…
MPLS, and the full Pensando team certainly have a mountain ahead of them. But I think they have the equipment they need to find purchase in their first tight niche and keep on climbing from there. I’m really excited to see how this plays out!