“Clutter and confusion are failures of design, not attributes of information.” ~ Edward Tuft
In the ever-so-neatly packaged and marketed buzzword bingo world that we know as “cloud,” there are two generally accepted flavors: on-premise, or private, cloud, and public cloud. Public clouds of the most well-known variety are Amazon Web Services (AWS,) and Microsoft Azure. On the private side you have VMware, if you buy a few-bajillion different software packages, or OpenStack, itself a fairly unwieldy beast of an ecosystem. All achieve roughly the same goals and end-state, namely allowing fast and easy creation and consumption of largely transient, virtualized workloads. The benefits and drawbacks of each, however, exist in different spheres.
The public cloud providers offer very easy to consume services, already built in their environments and on their hardware, as secure as we can call anything these days, all for an ostensibly nominal fee. Your data lives outside your data center, however, and can suffer from what’s known in the industry as “noisy neighbor” syndrome, whereby other users’ applications hosted on the same hardware as yours, can consume enough resources to starve your applications. Additionally, as they say in the valley, AWS is very cheap if you fail, it’s very expensive if you succeed, meaning that the cost of public cloud looks great at first glance, but once you start consuming a lot of resources, your costs can quickly balloon to eye-watering levels.
The private cloud ecosystems aren’t without their own blemishes, though, and while offering a much lower operational cost on paper as you don’t have ongoing fees to a provider, and you ultimately have more security and control since your data lives in your own data centers, they tend to be very challenging to stand up, and even more complex to maintain. Many companies have to increase hiring of specialized staff just to monitor and maintain the system, and often have to pay consultants to get the environment built and tuned in the first place.
A visual example from a deck they presented on during a recent Tech Field Day event shows a traditional model of cloud, and the differences at a high level between public and private implementations of the same:
From what I can tell, ZeroStack seems to have been founded on the premise that OpenStack—one of the most talked about private cloud systems out there, and by far the one with the most buzz—is an incredible product, but needlessly complicated for all but a few folks. Their core mission to simplify and reduce OpenStack to an easy to use platform, deployable in minutes rather than weeks, is one that will undoubtedly resonate with a significant segment of the IT population. In doing so, however, they run the risk of colliding head on with some of the big boys of industry.
What ZeroStack does is commoditize OpenStack into a hardware and software platform which can be deployed in under 15 minutes, a claim which I can verify first hand. They ship a box—based on merchant hardware—to your location, you plug it in, answer a few questions, it phones home, and within minutes you have a fully functioning OpenStack environment. As anyone who has deployed OpenStack, even in a basic development environment, can attest, it is very time-consuming and not entirely trivial to stand up in any kind of functioning manner—presumably the end goal.
ZeroStack’s hardware is custom built from off-the-shelf components, and comes in four different flavors depending on your needs. The servers can “stack” in a scale-out model, and use a distributed storage fabric, distributed management, and distributed SDN fabric, across all servers. This allows for large build outs, but perhaps more importantly, it allows for seamless host failure recovery through a leader election mechanism. The more interesting bit is how they handle management, and is where we begin to see where we can draw a Meraki comparison.
What ZeroStack has done that is the most evolutionary—I won’t say revolutionary since this is being done already, more on that in a minute—is moved the management components of OpenStack into a cloud (how many abstractions of cloud can we handle before the whole thing blows up in a cloud of vaguely consultant-smelling marketecture?) They host this on their own platform in their own data center. This allows you to manage the system from anywhere, much like what Meraki did for wireless networks.
This can be illustrated again with a slide from the same presentation referenced above:
By separating what we could loosely call the control plane from the data plane, to borrow from the networking world , the entirety of the OpenStack system and deployment model is made manifestly easier for the average entity to deploy and consume. You rack and stack the hardware, point it at the ZeroStack cloud management portal, and it does the rest. You get a cup of coffee and when you’re done you have an on-premise cloud. There are obviously some subtleties to the deployment, and extra knobs you can tweak if you choose, but this is far quicker than a traditional deployment, and should appeal to many people.
The main risk I see from a long term viability perspective is that this model of OpenStack deployment puts ZeroStack squarely in the path of at least Cisco and HP Enterprise, with their MetaPod and Helion CloudSystem Enterprise products, respectively, which perform almost entirely the same function at a higher cost point. Cisco’s solution, in particular, can be more accurately compared to a VCE vBlock than to the ZeroStack platform, the former coming pre-racked and plug-in-to-power ready with full-blown UCS compute, Nexus 9K switching, and ASR routing. Cisco’s solution has the further benefit of being fully deployed and managed by Cisco, and so is quite literally a plug-and-play solution. HP Enterprise’s Hellion CloudSystem utilizes VMware for the cloud platform, but functionally accomplishes the same goal of distilling what can be a complex deployment down to a single purchase proposition.
I think where ZeroStack has an advantage, and possibly an unchallenged market space niche, is in the lower to mid-tier price points. Many (many, many) companies who wish to deploy OpenStack simply won’t be able to afford the Cisco or HPE solutions, but still have a desire and a need for a simplified deployment model. If ZeroStack’s products can run in a stable manner as well as they can deploy, I think they have a fighting chance of remaining viable for some time to come. Either that or they’ll establish enough of a market footprint, and become enough of a challenge the the big-boys inevitable desire to sell down-stream in a likely scaled down product and price point, that they’ll be acquired for a surprisingly large amount of coinage—not at all out of place for the valley.