The Emergence of Datacenter-as-a-Service
Image: Depositphotos
Source: CogWorld Think Tank on VentureBeat
In my role as an investor, I’ve been hearing a lot of chatter lately about datacenter-as-a-service (DCaaS), the notion that datacenters should move towards a cloud-native architecture delivered as a service and that enterprises should no longer need their own IT ops teams. DCaaS promises to separate a company not only from its physical infrastructure but from the notion of physical infrastructure itself.
It’s an appealing idea. Two decades ago, Amazon, with the massive compute infrastructure it had built for its ecommerce operations, realized it was sitting on a gold mine: It could “rent” out its unused capacity for companies that would rather create a server with a few commands (or mouse clicks) than by “racking and stacking” physical hardware. Amazon Web Services was born. But as front-end apps become even more accessible — and commonplace — it’s more important for the back-end infrastructure to become as agile as those front-end apps. What good does it do to be able to build your company on rented hardware if your entire business goes down in an AWS outage?
DCaaS is a bid to solve that dependency. But while it’s tempting to think that moving to DCaaS is as simple as moving your infrastructure to a public cloud provider, it’s more complex than that.
Datacenter-as-a-service involves treating the entire datacenter with the same cloud native care with which you treat your applications. With an application, you assume that the container, or even the server on which an instance is running can go away at any time, and you architect your application to survive that using an orchestrator such as Kubernetes, which maintains multiple instances and moves them to other resources as required.
Datacenter-as-service takes that way of thinking all the way up to the datacenter level, providing multiple options so that if one datacenter or computing resource goes down, you can move not just your front end, but your back end applications to another datacenter or computing resource.
But as important as it is to keep applications running, having the option to use multiple data centers on command offers additional advantages, such as:
The ability to make use of private cloud resources in a truly hybrid way, such as keeping sensitive back-end data on private cloud servers while using public cloud for less sensitive front end applications
Strategically locating data centers inside (or outside) of countries with specific regulatory requirements
Following the notion of “data gravity,” in which applications that process data are deployed close to the data itself
Enabling the dynamic creation of edge computing clusters in data centers that are most appropriate, including on non-traditional devices.
While a public cloud like AWS can certainly act as the base datacenter behind this functionality, true DCaaS provides an additional level of orchestration. Combining that coordination with the advantages of the cloud-native datacenter creates a powerful combination of agility and cloud that can serve developer and architect needs in a way we haven’t seen before, enabling innovation we can only begin to imagine.
The extent to which a company provides DCaaS varies. For example, some companies, such as the major hyperscalers (AWS and Azure) and smaller players such as vXchnge and CoLocation America provide what is essentially managed hosting, enabling the remote creation of data centers, but limiting those data centers to a single provider.
On the other end of the spectrum, DCaaS provides a combination of software and services to provide a seamless managed service. Mirantis, for example, provides a fairly complete implementation for this with its Mirantis Flow offering. The company provides a dashboard from which to manage new clusters for bare metal, virtualized, and containerized applications on both private and public clouds, as well as 24x7x365 support and a pay-as-you-go model. It promises an unprecedented level of elasticity for computing resources.
Datacenter provider Equinix takes a middle ground stance, providing managed hosting and other similar DCaaS services, but also partnering with Mirantis so that customers can deploy bare-metal clusters on its infrastructure. (Full disclosure: I have no investments in any of the companies mentioned in this story.)
We’ve moved from a world in which humans came to applications to one in which applications come to humans. They come in the form of embedded devices such as medical equipment, appliances, and cars, and in the form of almost constant connectedness. It is said that 2.5 quintillion bytes of data are created every day, but that number has been floating around since at least 2018 — and we’ve added more than 600 million new users to the internet since then. To keep up with that kind of demand, the datacenter needs to be as agile as possible, and DCaaS is one way — and perhaps the only way — to make that happen.
Sanjit Singh Dang is Chairman and Co-Founder at U First Capital. He is also a member of the Cognitive World Think Tank on enterprise AI.