By Jeff Paine, VP of Marketing at Pica8

The Battle for Network Supremacy Shifts to the Enterprise

The data center networking-vendor/architecture wars are officially over.  All seven of the largest web-scale companies, representing about half the total market, settled on white box switches and disaggregated Linux-based software for their core network infrastructure, leaving the traditional vendors to fight a holding action amongst themselves for the remaining market share.   

In 2020, all network-vendors' eyes turn to the largest networking prize of all -- $8B to $12B, depending on who you ask -- the enterprise, and, more specifically, to the campus and access networks of the world, most of which are long overdue for modernization. (Some have still barely been touched since the last big upgrade cycle prepping for Y2K.) 

In a normal year, Cisco's decades-long stranglehold on the enterprise networking market would protect them from unwanted competitive attention in this, their home turf, but forces have conspired to make 2020 the year that enterprise networks are finally in play, in some cases perceptually, but, in other cases, we could be seeing real impact and the early stages of Cisco Colony Collapse.  

So, what, explicitly, do we expect to see happening in 2020?

* "Traditional" networking vendors get noisy - Cognitive Campus!, AI-Driven Enterprise!, and so on - but fail because they lose the thread that the vast majority of enterprises are - above all -- desperately seeking operational simplicity because of the extreme shortage of - very expensive - networking engineering personnel.  To date, Cisco's "big iron" competitors haven't found much success in this market, and their new expansionist approaches that require the minting of even more equivalents to pricey CCIEs - increasingly known as Cisco Crash Investigation Engineers - to run these new architectures seems quite counterintuitive, actually.

History also suggests a second reason they will fail -- "proprietary is dead, Jim."  At the most basic of levels, if enterprises felt they could "escape" Cisco vendor-lock by trading for another flavor of vendor-lock, they would have pressed that button years ago. 

Finally, legacy vendors also crash headfirst into the "to a man with a hammer, everything looks like a nail" problem.  What works for them in the hothouse environment of the data center - automation, AI, DevOps workflows - does not apply to the organic chaos of bespoke enterprise network topologies and workflows. That doesn't mean the legacy vendors won't try, but automation and, to a lesser extent, AI, require very different implementations in enterprise campus and access networks.  Also, in the data center world, IT = revenue, but in the enterprise, IT = cost, which also requires a very different mindset.

* In the enterprise, Open Networking finally evolves from "NOS-on-a-Box" products to "complete white-and brite-box access and campus network solutions," finally cracking open "Fortress Cisco" in large enterprise accounts.

Cisco's previous enterprise "immunity" to predation from the white box "movement" that has been plaguing them - and others - in the data center largely depended on two critical things not happening in the market - 1) the development of an open networking, Cisco DNA Center-class automation framework for campus and access switches, and 2) the availability of a broad palette of open switches from a trusted global brand with a proven supply chain and worldwide support on par with Cisco.

As 2020 starts, both of these are now uncomfortable new realities confronting Cisco (and the other legacy networking vendors).  Pica8, for example, has introduced a Cisco DNA Center-class enterprise automation framework for open campus and access switches that allows even non-programmers to turn on and configure thousands of switches; and Dell EMC now has the broadest range of open access and campus switches ever seen in the branded white box market. All these switches can run Pica8 software and be supported by Dell.

* 2020 will also see the start of a debate on what should enterprise network AI look like and what platform it should ultimately run on.  Replicating the firehose of telemetry ingestion model from the data center for enterprise networking environments appears problematic and will probably give way to a more curated, on-demand telemetry ingestion approach, perhaps using a technology like GraphQL.

On the platform side, pushing all decision-making off to the cloud is starting to feel a bit like some sort of modern variant of IBM 3270 emulation, but some of the real-time security and analytics decision-making has to take place directly at the access edge of the network, meaning that much of the policy intelligence is likely to ultimately sit either on the access switches themselves or, possibly, on APs.  House money is on access switches, as they are far, far "beefier" platforms than AP hardware, with some access switches already sporting 4-core processors.  When that happens, it's likely that WiFi simply becomes the "last mile" to the access switch ports, and AP hardware slims down even more for cost savings.

##

About the Author

Jeff Paine leads all marketing strategy and execution for Pica8. He has 25+ years of marketing experience in the networking industry, starting his career as the very first (pre-IPO) marketing hire at Cisco, where he successfully popularized the then-arcane technology known as routing by inventing both the router icon and the brand iOS. Post Cisco Jeff has run marketing and/or business development for large public companies, such as UTStarcom, and has executed the marketing launches of a number of networking/security software companies, including NetCitadel and Tigera.

Jeff holds a BA in Communications from Brown University and an MA from the University of Texas at Austin. Jeff also earned a minor in Ethnomusicology from Brown and played in a Balinese gamelan orchestra for two years as well learning to play the Japanese koto.