Juniper’s software play, why you’ll benefit even if you never buy it.
Juniper made a bold move this week, announcing software disaggregation on their switching products. They are fully committed to allow Junos to run on other vendor’s switches, and allow other vendor’s OS to run on their switches. This will start with the new QFX5200 and fully transition in the next 12 months.
We’ve seen this coming for a while now, with merchant silicon, whitebox hardware, an some smaller players already taking smaller steps toward this separation. But Juniper is one of the big incumbents, and already has a big stake in the market. Their switching sales have been taking marketshare the past few quarters and have been one of the early bright spots in what appears to be a surge in enterprise sales for them.
Now, for myself, I am an enterprise customer. I already buy lots of Juniper gear, and I already understand the value of Junos to me. As the decision maker in my company, I don’t see myself buying any disaggregated hardware in the next 3-5 years. I’ll buy 1st party hardware and bundled software because that still makes sense for my organization. And yet I see tremendous value in this move, and I think customers like us will still benefit.
The two takeaways for myself are:
Its the software stupid! And now I’m certain Juniper understands the value of Junos. Let’s be honest, most switches today are almost chip for chip the same. For the past 3 or 4 iterations we’re all buying the same Broadcom silicon, with different paint jobs on the tin. Sure there is some customization, but the core hardware features, speeds and density are all dictated by the chipsets and so there is little value add between the vendors on the hardware. So what make Juniper, Arista, Cisco et all different? Its the software. And Juniper has had one of the most powerful and flexible software stacks in the industry. They have best in class automation, open standards and interoperability, and a CLI that users fall in love with. And with disaggregation, their developers are free to focus on core features and iterate more quickly. I also believe this will lead to fewer versions of code paths for them to support. The abstraction layer they’ve built should allow the OS to be more similar across all platforms and features.
Abstractions are awesome! This disaggregation didn’t come out of thin air, it has been a 3 year effort that has already offered benefits to customers and Juniper. First a summary of the tech. Juniper has built a new Linux foundation that powers the hardware of their products. The traditional Junos now runs as a VM on that Linux OS. They’ve already shipped several successful products on that foundation. With that Linux foundation and abstraction we’ve seen some great new features. The vSRX and vMX have been possible, largely because of that abstraction. The QFX5100 can run In-Service upgrades because multiple Junos VMs can simulate dual routing engines. And customers can also deploy VMs directly on the QFX switches, that run alongside of Junos and have direct access to the hardware and stats. Examples of that would be HFT apps, wireshark, or ELK stacks, running right on the switch. Juniper is also able to support new hardware platforms with only changes to the Linux stack, no code changes to Junos, no regression testing on their large code base. This means fewer bugs and faster support for the latest chipsets.
My take as a customer
This isn’t really a surprise move. Analysts and pundits have been talking about the rise of white box for several years now. We’ve seen smaller players take some steps in this direction, though none have gone this far. And though I don’t see myself purchasing gear this way, I love a.) the virtual editions of their products, b.) seeing new hardware come to market quickly, c.) the promise of fewer bugs and d.) Junipers commitment to Junos and the things that really make it special in the industry.