There are two fascinating and completely contradictory trends going on in the infrastructure space. On the one hand, we have the rise of so called "server-less infrastructures," the event-driven approach, first introduced by Amazon Web Services with its Lambda offering and since emulated by a number of public cloud providers (Google, IBM and Amazon for a start) as well as a number of initiatives.
The idea of these event-driven approaches is that processes sit idle, essentially doing nothing until a particular event occurs, at which time a process kicks off. Once that process is complete, the function shuts down and goes back into its vegetative state. The promise of these event-driven approaches is pretty exciting: All of a sudden developers and operations teams don't need to think about servers, and can instead focus on the stuff that matters -- building cool applications.
Of course, there is an entire world of organizations that aren't able to adopt these new approaches anytime soon -- either because of the regulatory environment within which they work, or some particular operating window that makes it unworkable, or perhaps their applications just aren't built in a way to leverage these new approaches. For these organizations, the promise of the cloud is a seemingly unattainable thing.
Which is where cloud appliances come in. An appliance is, as the name implies, a physical box that is pre-built with all the software and systems needed to make it run.
Perhaps the best way to understand what an appliance does is to think about Google's (ironically recently discontinued) search appliance. With the Google Search Appliance, organizations got a physical box that, locked tightly within many layers of physical and virtual security, contained Google's search software. Organizations placed said box within their organization, after which the GSA began to crawl and explore the organization's data to create a "Google-like" search offering of all the data behind the firewall.
An appliance approach is something that arguably lends itself well to OpenStack, the open-source cloud computing solution. Since OpenStack is all about software that people can use in ways that the best work for them (as opposed to a proprietary public cloud offering from the likes of Google or AWS, which is much more of a "take it or leave it" deal), it makes sense that vendors would stand up to offer potential customers OpenStack in every flavor under the sun.
Indeed, one of the criticisms of OpenStack has been that it is just too fragmented. In its defense, the OpenStack community has long been trying to resolve this issue and create more consistency between the different offerings.
Anyway, back to appliances. With the OpenStack Summit in Austin less than a week away, the announcements are coming thick and fast and Mirantis, a company that has made headlines around the OpenStack community for years, is standing up and finding some new appliance love. The company is partnering with Supermicro and Arista to offer a certified appliance for both public and private clouds. At the same time, it's partnering with Quanta Cloud Technology and Cumulus Networks to offer essentially the same thing.
Some from outside the industry might find it a little unsettling to see two press releases from one company announcing near-identical partnerships with two competitors on he same day. But in the technology industry . . . anything goes.
Anyway, it is interesting to see Mirantis position this move -- bearing in mind that it sells OpenStack through a number of different models. As Jim Sangster, Mirantis' senior director of solutions marketing, points out, while converged infrastructure (the generic term for anything that involved hardware plus software) may sound like a less than cool way of doing technology, companies spend around 20%-25% of their IT infrastructure budgets -- billions of dollars per year -- on converged infrastructure.
As the world moves to the cloud, Mirantis' belief is that there is a similar market opportunity for OpenStack. Sangster predicts that consumers of OpenStack will adopt the same spending pattern, with 20%-25% of OpenStack budgets being spent on OpenStack as converged infrastructure.
While that converged infrastructure move flies in the face of the promise of our server-less future, Sangster posits that the value that converged infrastructure delivers -- by taking a group of technologies that can be difficult to use on their own (much less together) and combining them into a prescriptive, pre-integrated solution -- is eternally attractive.
Sangster points out that OpenStack has, until recently, been viewed as software for innovators and early adopters. This is the realm of proud DIYers blazing the trail ahead. They love to experiment, doing all the hardware and software engineering possible as they work to understand, implement and eventually deploy a new system like OpenStack. This is, of course, fun for the tinkerers, but unhelpful for the mainstream organizations that simply want to use a solution. For those folks, converged infrastructure makes sense.
Or it would, were it not for the fact that people have tried this before. Indeed, well-funded startup Nebula was founded to create an OpenStack appliance -- and then shut up shop due to the fact that it couldn't find traction.
Sangster heads off this argument. "Today, adoption of OpenStack is crossing the chasm to the early majority part of Moore's curve, where proven yet innovative solutions providing low risk prevail," he says. "The OpenStack User Survey showed that standardization was the most frequently chosen business driver for choosing OpenStack. Meanwhile, there has been an increase in the volume of companies using OpenStack, and in the proportion of them doing so in production -- again signaling that the early majority is arriving for OpenStack. OpenStack's early majority wants easy-to-use, easy-to-adopt, pre-integrated solutions."
Hence the timely (in Mirantis' eyes) introduction of these new appliances.
MyPOV
Horses for courses. At a purely conceptual view, AWS Lambda and its competitive event-driven offerings are the best way forward. But concepts often have a tendency to fail when applied with a broad brush to the general mass of organizations, which is where a solution that doesn't perhaps have all the early adopter credibility, but is fundamentally easier to use, comes in.
That is the real benefit of appliances. While they may see reduced scores in the race for bleeding-edge cred, they more than make up for it in their general applicability. Mirantis isn't going to stop offering its standard software-based product anytime soon. But by creating a channel it offers up not only a new consumption model but, even more importantly, another route to market.