It’s been a while since my last post (months) so I’m trying to ease back into the saddle with a topic that everyone can appreciate: the Mexican Standoff.
Before I get into the topic, in the interest of full disclosure readers should know I’m the networking co-chair at the Open Compute Project and that I work for Cumulus Networks in my day job. The opinions in this post are all mine, though.
And off we go! For those of y’all not familiar with the concept of the Mexican Standoff, Wikipedia offers this definition:
a confrontation between two or more parties in which no participant can proceed or retreat without being exposed to danger. As a result, all participants need to maintain the strategic tension, which remains unresolved until some outside event makes it possible to resolve it.
For the visual folks out there, here’s a Mexican Standoff in film:
So how does this relate the networking industry? I’ll describe a situation that’s being played out today.
You have a merchant silicon company, Broadcom, that dominates the market. That company is then bought (“merged”) by another merchant silicon company, Avago, that does custom ASIC work for a variety of folks.
On the networking side you have Cisco, the industry’s largest traditional networking vendor, which touts its custom ASIC chops along with its merchant silicon boxes. Both chips come from the same vendor now.
Then you have Aristsa, another traditional networking vendor, that uses the same merchant silicon and loves to tout its ‘h-yuge’ buffers.
Neither of these vendors can ‘out maneuver’ the other with another silicon vendor without pressure being applied to their own bottom lines by their existing silicon vendor.
Welcome to the status quo in networking: essentially a Mexican Standoff.
How about another recent event?
A few months ago, I blogged about how the Open Compute Project has entered the Campus, Branch, and Wireless (I’ll call it CBW, that’s CB Dub-ya) market with key contributions to access switches and wireless APs. Since then, the CBW group has been heads down spec’ing and coding an OpenWRT-based NOS for these APs.
However, they hit a speed bump early in their development: because the APs are Broadcom-based, they used the Broadcom bootloader and of course, it requires the use of the Broadcom SDK to use the radios and the Gigabit phys. *ugh*
If y’all don’t know why I’m sighing, read this article, in particular the part about obtaining the Broadcom SDK for wired networks. This also applies to the wireless side of the house.
Luckily, after some prodding, the bootloader was changed from Broadcom’s proprietary one to UBoot. It didn’t solve the problem entirely, but after some further prodding, the team got what they needed (mostly) and is marching along.
Unfortunately, the need to wrestle with large incumbents is normal for the industry.
Why Is The Networking Industry Like This?
Well, it wasn’t always like this; in fact, this situation arose over the last couple of decades. Lemme explain.
For the longest time, there was no competition in the market. Every vendor had its own kingdom and alliances (7 kingdoms anyone?), and when a new challenger arose to threaten the throne of one or more of these houses, an unlikely alliance was made and the potential usurper was put down.
How? Essentially, the technology was replicated by a vendor and its product/sales team, pushing the usurper out of the market; or was brought into the fold by acquisition, merger, whatever. No more threat to the crown.
Sometimes (if not most times) these acqui-mergers were great for investors and shareholders, but not customers or the industry.
How Do We Change This Mentality?
When I say ‘we’ I mean the ‘royal we’.
The solution is as simple as voting with your dollars. Back when I was at UTSA (before I was the OCP Networking co-chair), I wrote a post on the Democratization of the Network Beyond a Two Party System. The post highlights how Open Networking resolves the Mexican Standoff by enabling you to use the hardware you want, the software you want, the configuration management you want, with the ability to change any piece without lock-in.
For too long, we felt too much like this when choosing our network:
No more. In Open Networking, it’s no longer a two-party system. It’s a fully representative populist one. Will the system change? There’s a reason it’s no longer called Slaver’s Bay (it’s now Dragon’s Bay).