This is one of those things I’m not entirely supposed to say out loud given my $dayjob, but I’ve been watching the industry back-and-forth since VMworld with bemusement and more than a few eyerolls.
So I was at VMworld at the end of August, and basically it was NSXworld for me. I took in the keynotes, Casado’s Spotlight session, and assorted “SDDC architecture with NSX” pitches. And of course I’ve been following the subsequent online slapfest between assorted VMware and Cisco personalities.
Exceedingly dignified soldiers
What I took away from my survey of the NSX talks at VMworld was that the relationship of overlay and physical networks has been transmitted within VMware via the old game of Telephone.
Casado was reasonably nuanced in his talk, acknowledging the continued necessity of physical networks, highlighting the need for highly reliable, low-latency forwarding and so on. He just focused much more insistently and at greater length on the problems he believes network virtualization solves, and could solve in the future. There were a couple of slides that attempted to address networkers’ concerns (P-V visibility, for example), but he breezed over them pretty quickly. I found his arguments somewhat unsatisfactory and remain unsold, but given the relatively short shrift they got I was also left wondering if the slides were inserted by others around him who wanted him to address common objections for them.
Sessions given by more junior PMs were another story. There were consistent claims that you could do everything you wanted to do with regard to your VM connectivity from within NSX without ever worrying about the physical network or as one said explicitly, “Without ever having to deal with the networking guys.” And not just VM-to-VM connectivity, but also connectivity for physical workloads. (Yes, there are ways to make this work—with cooperation from network hardware partners.) Problem is, this also seems an unnecessary land grab, especially from the point of view of actual networking professionals, who already get a relatively small slice of the overall IT budget and have their own stories about things they wind up having to do to fix VM networking issues.
Moving further down the line, we get the general mass-marketing and sales version of the story, which of necessity was distilled down to a few key talking points. Nuance inevitably gets lost in this process, since in order to get everyone to digest the main point you have to make it as simple as possible, and then tell them to go forth and replicate it.
Then there were the keynotes. Senior-level executives typically get an even more concise briefing of the key points of an issue than sales and marketing teams do. In addition, VMware’s core business is steadily eroding from a market share perspective (albeit still with plenty of headroom in absolute terms). Heterogeneous hypervisor environments are becoming increasingly common. In other words, the senior executive team is under a lot of pressure to find new areas for growth. A superficial reading of NSX might appear to offer that potential, which may explain some of Mr. Gelsinger’s feather-ruffling comments during his talk.
So here’s the real challenge: VMware is in the midst of a mid-life crisis. It’s been extraordinarily successful up to this point, but now some are finding it a bit stodgy and are looking at younger alternatives. Such moments cause many people to behave in unusual and even self-destructive ways until they figure out what they want their second act to be about. This is perhaps even more true of organizations, which are made up of lots of people with a variety of different business and personal agendas.
As long as you still look bad-ass, age is just a number. Right?
I’ve been asked a few times recently whether every networking vendor should consider VMware a competitor. My personal belief is that’s really up to VMware to decide. There’s no technical reason that should be so: overlay networks absolutely rely on the physical networks they traverse and at the same time, address different needs. The relationship is both symbiotic and complementary, as is the expertise of the respective IT groups (and vendors) involved. All in all, either/or positioning (“Software solves everything!” “Hardware for the win!”) seems a complete distraction to me, as it diverts energy from helping users draw circles around what functionality would work best for them within their physical network vs what overlay networks need to be optimized to do.
All that said, it’s clear that certain groups or individuals within VMware see good business reasons for setting up such a dichotomy. It’s not irrational, since Tech as a whole is becoming middle-aged and increasingly oligarchic, with big players of all flavors increasingly trying to offer the whole stack to their customers (though I would think that the woeful state of HP’s overall health would offer some pause to those trying to emulate them). Unfortunately, it’s rare that oligarchies are all that great for consumers, and market stability rarely lasts more than a decade in any case before someone else figures out how to do it better/faster/cheaper (cf Detroit vs Japan).
Given the apparent diversity of postures vis à vis physical networks within VMware—and the irrefutable dependency VMware has on hardware partners to deal with physical workloads–I’m not ready to nod along with Cisco friends who feel VMware has unequivocally declared war on the entire networking sector. There are unquestionably messianic types on both sides, but also enough realists that I’m willing to hold my breath and wait out the near-term competitive euphoria, to see who ultimately wins out.