I have been spending a lot of time investigating the OpenStack ecosystem recently, including a journey to Hong Kong for the OpenStack Summit. In this post, I intend to present some thoughts and observations on the state of OpenStack and why the momentum it has gained recently could mark the cusp of a very important transition in the IT industry. I don’t intend to describe what OpenStack is, so check out http://www.openstack.org/ if you are new to the game.
Here is the TL;DR summary:
1. OpenStack has picked up a lot of momentum in the last 6 – 9 months. This is significant because if the momentum continues, OpenStack could prove to be a classic disruptive technology, the kind of disruptive force that makes venture capitalists excited and incumbent vendors fearful.
2. OpenStack and truly dynamic cloud architectures represent a fundamental rethinking of IT infrastructure, not just an evolution of existing static virtualization infrastructure.
3. How effectively VMware and other incumbents navigate this transition is an important open question in the industry. My expectation is the cracks in VMware’s armor will become increasingly apparent moving forward.
4. There will be many opportunities for new and innovative vendors to capitalize on OpenStack and the shift to dynamic cloud architectures. Thinking futuristically, I am interested in the potential for new technologies to replace today’s hypervisors and perhaps even operating systems.
The Cloud Shift
OpenStack has gained a lot of steam in the last 6 – 9 months due to a confluence of growing support from heavyweight IT vendors, crossing the “production grade” maturity threshold, and the growing competitive pressures on certain classes of enterprise and service provider to rethink their infrastructure capabilities. If OpenStack is a long-term success, we will likely look back on this period as the key inflection point. It is still early days. When I first started trying to estimate the scope of existing OpenStack deployments, some people told me there were perhaps 10k – 20k physical servers in production. I think the actual number is at least 40k, but this number is still not huge in the grand scheme of things. More significant is the number of PoCs underway, including some in mainstream enterprises like insurance companies rather than just early adopters like service providers. The OpenStack professional services firms are going gangbusters as the market tries to figure out how to get started with this technology (alluding to one of the near-term challenges: it’s too hard to deploy).
The cloud architecture embodied by OpenStack represents a fundamental rethinking of IT infrastructure, even though some CIOs still fantasize that the cloud is merely an evolution of their existing virtualization infrastructure, and some vendors still fantasize that they can get by with merely “cloudwashing” their existing products. Amazon Web Services pioneered how a true cloud infrastructure should behave: completely self-service, rapidly and automatically provisioned, elastically scalable. They also set an example for how this infrastructure should be implemented: with pools of commodity hardware managed by powerful distributed software (not that other web-scale companies like Google and Yahoo weren’t exploiting this model in some fashion as well). The fact that such an architecture makes heavy use of server virtualization is just about the only major similarity with existing enterprise IT infrastructure. In a true cloud architecture virtual machines are short-lived, run on cheap commodity hardware, and are highly tolerant of the failure of any individual node. In a traditional static virtualization environment, VMs are long-lived, run on expensive dedicated hardware, and have the expectation that their dedicated hardware will be nursed back to health if it goes down.
More enterprises and service providers are waking up to the competitive reality that they need to either use AWS or make their infrastructure work like AWS. Not for cost reasons, but for revenue reasons – they need to launch new services rapidly to remain competitive. OpenStack is a way to accomplish the necessary infrastructure transformation. It isn’t easy, and not just because the technology is new, but also because such an architecture requires new organizational processes, and people are slower to change than technology. Nevertheless, there are many signs that the cloud model is the future of infrastructure, both public and private.
The Incumbents’ Dilemma
If this is the case, a billion dollar question is: how will the incumbent vendors navigate this transition? Many large incumbents have something to lose here, such as those who have built their businesses selling big centralized hardware products that are now threatened by the paradigm of commodity hardware pools. But the one I am really watching is VMware. The data center is becoming software-defined, and VMware is a software company which would seem to position it more favorably than hardware incumbents. VMware has a very dominant position in server virtualization, which it is wisely leveraging to become more entrenched in business processes. But again, virtualization is just about the only similarity between a traditional enterprise VMware environment and a true cloud. Neither VMware’s technology nor its economic model is geared toward a massive number of short-lived VMs running on pools of expendable commodity hardware. Its pricing model doesn’t work in this new paradigm and the escalating costs of running a VMware stack are causing a number of enterprises to start at least evaluating alternatives to avoid lock-in.
In many ways, OpenStack is a classic disruptive technology to VMware: it starts off inferior on many axes that matter to the mainstream, and VMware’s bread and butter accounts won’t touch it with a 10-foot pole, yet with time it could become superior to VMware even on the traditional axes. There is still growth for VMware at the high end of the market, where the most mission critical workloads have yet to be virtualized, and accordingly VMware is putting much of its energy into tapping this remaining bastion of growth. Yet concurrently, the low end workloads like test & dev (where VMware itself got its start) are starting to be placed on OpenStack-based clouds. As VMware moves up market, the low end is being eaten out from under it in a classic disruptive wave. We are starting to see some organizations bifurcate their infrastructures, keeping legacy static workloads on a traditional stack powered by incumbent vendors, and placing new dynamic workloads on clouds powered by OpenStack and new scale-out technologies like Ceph. With more maturity, OpenStack could eventually become superior to the legacy stack even for mission critical workloads, and the disruptive cycle will come full circle.
Does VMware comprehend the risk, and understand what it must do to adapt? My sense is the leadership at VMware gets it. They are smart people. The acquisitions of Nicira and DynamicOps were in some ways an acknowledgement of the changing landscape. But translating acknowledgement into execution all the way down through the organization is very hard. How many big companies are able to successfully navigate such a major transition and come out ahead, as opposed to simply preserving the legacy business for as long as possible while ever-so-slowly being supplanted? How many incumbents successfully navigated the smartphone and tablet revolution, for example? My expectation is the cracks in VMware’s armor will become increasingly apparent moving forward.
There are many risks to OpenStack, such as its decentralized structure as an open source project and the current tendency to value new features over unglamorous things like stability. However, whether or not it ends up being OpenStack, some form of cloud management system will ultimately make a big mark on the industry, and right now OpenStack is the clear lead contender.
This is an exciting time for OpenStack and the cloud. There will be many opportunities for new and innovative vendors to capitalize on the paradigm shift that the cloud represents. One area I think is particularly interesting is the emergence of new computing models that replace virtual machines entirely. We are already seeing a lot of interest in things like Docker and various other containerization approaches. Recently Rackspace acquired ZeroVM, which creates distributed VMs that can tap into massive distributed processing power rather than that of a single host. Moving yet further down the stack, not even Linux is safe in this new world, with cloud-optimized operating systems like OSv emerging. These ideas could be the topic of entire blog posts in themselves, so I will leave things here. I look forward to discussing these ideas further with anyone interested.
Thanks for reading!