A recent Gartner blog calls the open-source SONiC network OS a disruptor, and predicts that by 2025, 40% of large data center networks will run SONiC in production. Gartner qualifies large DCs as 200 or more switches, which to my mind means very large enterprises, service providers and telcos, and cloud and hyperscale giants.
The Gartner blog predicts that “…during the next three to six years, SONiC will become analogous to Linux as a server OS, allowing enterprises to standardize on an NOS that is supported across hardware vendors.” Gartner also notes that client interest in SONiC jumped 87% from 2020 to 2021.
Why the increase? It’s possible that SONiC is just the enterprise version of cultivating your own sourdough: a homesteading instinct spurred by pandemic isolation. But Gartner is betting this is more than just an Instagram craze. My take is that SONiC needs three things to gain significant traction in the enterprise:
1. Commercial support, a la Red Hat
A commercial version backed by a vendor can provide technical support, software patches, updates, and so on. A commercial version should also handle the fiddly bits with the ASIC abstraction layer (the SAI) so enterprises don’t have to fuss about the hardware–give the engineers a list of switch platforms on which SONiC runs out of the box.
2. Commercial management, configuration & automation tools
SONiC is Linux at its core so enterprises can manage it with tools such as Chef, Puppet, and Ansible, but network engineers may want something more tailored to their discipline. Apstra is making a bid here, but you also have to buy into Apstra’s entire worldview, which includes embracing Intent-Based Networking as an operational model and deploying hardware and cabling to Apstra’s specifications. There’s room for Meraki-like approach to managing SONiC—i.e. stripped down, simplified, and cloud-based.
3. Sales & marketing to promote disaggregation in general & this NOS in particular
Hyperscalers and big cloud companies have a clear business case for separating hardware and software, and for embracing a network OS with a modular design that’s highly customizable. Most enterprise shops don’t care about these attributes. They want reliable software; support for core protocols; and tools to configure, manage, and troubleshoot the network–ideally delivered by a vendor partner.
Is There An Enterprise Case For SONiC?
As you may know, SONiC was originally developed by Microsoft for Azure before Microsoft released it as open source software within the Open Compute Project. SONiC’s origin story is one of a hyperscale NOS that now walks among enterprise mortals.
Can SONiC succeed where other enterprise NOSs have yet to make a significant dent in the market?
Well, many networking vendors with their own OSs to peddle also support SONiC, including Juniper Networks, Arista Networks, and Cisco. If you want their hardware and the SONiC NOS, they will sell it. I suspect that’s driven more by cloud buyers than the typical enterprise customer, but it’s there for the asking.
In addition, Dell Technologies has released its own distribution targeted at the enterprise, the aptly-named Enterprise SONiC Distribution by Dell Technologies.
SONiC also has an advantage that other network OSs don’t have–Azure. ASIC makers and switch vendors alike want to sell to Azure, so the likelihood of them playing funny games with things like ASIC SDKs and the SONiC Switch Abstraction Interface (SAI) is unlikely (not impossible, but unlikely). Microsoft can also throw development resources at SONiC to ensure a base level of viability for the open-source code.
That said, disaggregation–separating network hardware and software–has not caught on significantly in the enterprise. It’s true that a disaggregation ecosystem has emerged, with ODMs such as EdgeCore and others providing switch hardware with Broadcom ASICs, and software companies such as Cumulus and Pica8 offering network OSs. Now SONiC joins the mix.
But what are the benefits of disaggregation for the enterprise? The typical story is outlined thusly:
- Independent advances in software and hardware
- Greater choice of hardware and competitive pricing on switches that use the same Broadcom ASICs as your favorite legacy vendor
- The ability to swap out your hardware or your software based on changing requirements
- Slightly more leverage in negotiating with legacy vendors
Are those enough? My impression is that without the three elements I outlined above, SONiC won’t make much of an enterprise dent. While Dell Technologies is making a stab at my list, it has more work to do, particularly on points 2 and 3.
I wonder if Gartner’s bullishness will serve as a signal to an ambitious startup or to traditional network vendors to take SONiC and run with it in the enterprise. We can check back in 2025 and see where SONiC stands.
