Is the Nicira acquisition good for OpenStack? Definitely. But before I defend my assertion, let’s look at why this is good news.
I’ll start with Microsoft (though the old-school tech guys who need to pay attention are at Cisco; more on that in a moment). Microsoft don’t get no respect in the infrastructure business — not least when it comes to marketing. What gave me hope was their truly funny campaign at vm-limited.com, with fabulous send ups like “The Hidden Bummers of Going Too Far beyond Virtualization.”
You may have seen this, where they went after VMWare with the 80s retread ‘Tad’ blathering on, under what was once known as a cool haircut matching red hair and orange corduroy trousers. (Admit it, you know people who think this is really fashionable.) The IT past, just a bunch of servers and storage running applications. Yeah, those were the days.
$1.2B later, VMWare has had the last laugh.
VMware gets tremendous credit for making virtualization an absolute necessity for any enterprise infrastructure. They harvested huge efficiencies and profits by severing the 1-1 relationship between servers and operating systems. They introduced unheard of opportunities for accelerating IT cycle times. This limited kind of virtualization created the momentum behind what we now call cloud, feeding the appetite of new application development technologies, that in turn were better for chasing new market opportunities.
Except for networking. While you could spin up VMs and deliver storage resources with more and more agility, network configuration remained a ball and chain. It is slow, painful, error-prone, vendor-bound and opaque, rife with obscure protocols, arcane data models and esoteric authentication mechanisms — and that’s in a standard, old-school one-OS-per-CPU environment. Add the fluidity of virtualized hosts and storage, and you had a two-speed fabric. Automated Servers/Storage plus Fred Flintstone Network Management.
Just as increased compute power and storage density unlocked the potential of server virtualization, increased network power and bandwidth created the headroom for rethinking networking in ways unseen since the demise of token-ring over co-ax. And this is where Nicira really blew things wide open. Their basic concept of L2 over L3, with edge-level Open VSwitch as the management band handling VLAN reconfiguration for VMs moving from server node to server node, challenged some pretty well-worn network verities.
First, it made the network as straightforward to manage virtual machines and virtualized storage. Second, it asserted the primacy of the application-driven infrastructure, in which network resource allocation is driven from the compute side. All that expensive network hardware with its unique, differentiated routing logic and obscure control methods? “Shut up and give me wide-open flat L3, and don’t worry your pretty little routers with how VMs communicate. We’ll take care of the logic”. In fact, by making the network a logical entity like servers and storage, Nicira’s approach — built on Open VSwitch — cut off the ankle cuffs on the network ball-and-chain.
It’s a brilliant move by VMWare. With the trifecta of storage, server, and network virtualization, they are set to drive the same innovation in cloud computing that they did in ‘classic’ datacenter virtualization. Throw in the investments they’ve made in application development — both directly in SpringSource and indirectly (via EMC) in GreenPlum — they are riding the tsunami instead of wondering how come the beach suddenly got much bigger.
It’s also a great breakthrough for all the VMware customers who at one point bore an uncomfortable resemblance to Tad, the IT mustache-guy in in the Microsoft satire. Cloud is no longer a trend; it’s a fact of life.
For the OpenStack community, this is unmitigated good news. And it’s not because of VMWare’s assertion that this will drive more openness. Rather, the competitive jolt that this injects into cloud computing, to Amazon and to everyone else, will drive accelerated innovation. There is no serious alternative to cloud computing.
Will Nicira continue to be an important player in OpenStack, or will it become the catalyst to further VMWare lock-in of next-generation infrastructures? Too soon to say. The Nicira guys continue to enthuse about how open their roots are, but even with $1.2B reasons to get all the other players at VMware listening to them, it’ll be hard for them to get the virtualization giant to change course. SpringSource was much bigger, and VMware is not much more than als0-ran in open source application development.
My early prediction is that enough people have been burned by VMware lock-in that OpenStack will draw even more interest, since VMWare has sent a $1.2B message to anyone who questions the cloud model to get with the program. In the same way that Microsoft continues to command great loyalty from key segments of the IT marketplace, VMWare’s ‘all in’ virtualization will guarantee it a seat at the table for anyone building a more fluid, responsive infrastructure. And it guarantees a seat at the table for alternatives to VMware — not least OpenStack.
What about Cisco? Yesterday might have been either a really tough day. Or, if you read their press releases, it was one of blindered vindication for drinking their own kool-aid. I suspect it’s neither: they’re now all asking how to accelerate the programmability of their network fabric.
Of all the box-centric networking players, Cisco in particular has made the argument that taking the intelligence out of L3 routing gear is a mistake, on the grounds that flat networks don’t really work with more the common scenarios involving arbitrary network demand, security, compliance, and more. Will OpenFlow provide enough agility to those application developers and workloads that need assured network throughput? Necessary, but not sufficient.
If Cisco’s as smart as they’ve often proven to be, they’ll get much more aggressive to OpenStack to expose richer network capabilities, and tie their unique networking virtues more nimbly to compute and storage resources. If nothing else, the Nicira+VMware transaction raises the stakes for everyone in the cloud game. And OpenStack is in the right position to benefit.