Back to the future: the return of the on-premise data center?

Many businesses who had their own server rooms or data centers on-site often grew faster than they were able to expand their IT resources. If you couldn’t build infrastructure fast enough, cheap enough, or efficiently enough to handle all the servers you needed, collocation was the answer. It was flexible and gave you access to economies of scale that would have otherwise been unattainable. Unless of course you were Google, Apple, or Microsoft — then you just went and built something even more efficient than the what the colocation datacenters could. 

Mid-sized enterprises were still investing in IT hardware, but not in the networking, power, cooling, and security infrastructure it needs to operate. And with high density compute making lots of existing data center infrastructure obsolete at a rapid clip, it was a solution whose time had come. Scaling to meet demand got easier, faster, and a lot less risky.

Clouds on the horizon…

Then public cloud offerings such as Amazon Web Services began to appear. It’s a seductive concept: you pay only for the compute, storage, IOPS, etc. you actually use. Zero capex, 100% opex. That’s great if you’re a startup looking to expand rapidly, then possibly pivot into something completely different if things don’t work out as expected. But what if you’re one of those companies that built its own datacenter 10 years ago, and is now struggling to stay competitive while saddled with the inefficiencies of an obsolete facility?

Like fish out of water

Lots of our mid-sized enterprise customers who ran their own datacenters or used a colo provider have entire teams of IT professionals skilled in virtualization technologies such as VMware. But their AWS expertise is sometimes limited or non-existent. Unfortunately, migrating to the cloud is not as simple as just taking VMware Converter and converting your virtual machine images into Amazon instances. Unless a very fat bill from your cloud provider is not a problem, of course. You need people doing things like rightsizing instances, storage pools, and memory footprints to avoid skyrocketing service costs.

Therefore, these companies may have been able to avoid capital expenditures by switching to the cloud, but the operating expenses they incurred in some cases exceeded the capex of traditional options! You wouldn’t think it’s possible to get burned by a cloud, but we’ve seen it happen. After that type of experience, moving data and/or applications back to on-premise facilities retrofitted for higher densities suddenly looks attractive again. We can help here, with high-density backup power and monitoring systems to get the most out of the limited space in an existing facility. Of course, we are on the other end of it as well, helping to make the colocation facilities (where cloud services run!) as efficient as they are.

Performance issues vs. cost & scalability

Most companies that chose the cloud, however, actually felt it was a cost-saving move. For some, those savings came at the price of increased latency. One way to combat cloud performance issues is simply to book more resources from the cloud provider. But overprovisioning to get back to the same level of performance you had with colo or on-premise can eat up all your savings. :-/ The trick is to plan for just the right amount of capacity. But wait, wasn’t that the big disadvantages of an on-premise datacenter in the first place — if you guessed wrong on capacity, it cost you? So, where’s the advantage of the cloud here?

Correspondingly, businesses have in some cases chosen to return to a colocation or on-premise solution simply for performance reasons. In 2018, however, with lots of dark fiber around and colocation management software (cloud providers are frequently hosted in colocation data centers!) improving, those concerns have faded somewhat. Although some companies are in fact “moving back home” from the cloud, it’s undeniable that far more are headed the opposite direction. All you have to do is look at the robust growth of the cloud providers.

Mission critical data, compliance, and applications vs. data

Choosing between cloud, colocation, or on-premise solutions depends heavily on the specific use cases and applications. It can make more sense to have certain applications in the cloud. Especially when you’re talking about an edge computing scenario, for instance. Mission critical data, on the other hand, is a good candidate to host on-premise. There, you’ve got total control. And you’re not relying on the internet — an additional attack vector — to transmit that sensitive data.

Recent data protection regulations in Europe, for instance, require companies to know where their data is stored. When sensitive intellectual property or other data can’t be risked to a public infrastructure, on-premise is sometimes the only choice. With web services, you never know where you data is stored or processed.

Why hybrid solutions are our destiny

We can see that choosing cloud or colocation over your own on-premise data center is always going to involve certain trade-offs. Some applications and industries will always need to have servers and storage on-premise, under full control. And sometimes, you need a datacenter that can do exactly what you need it to do, and you can’t always get that from a colocation datacenter or web service.

Yet the advantages of the cloud are becoming greater day-by-day. Now we’ve got applications that have been designed for cloud computing from the ground up. And many companies are asking themselves, if we’re not in the business of running data centers, why should we be devoting so many resources to doing that instead of focusing on our core competence?

It’s difficult to see a way around either of these two arguments, which is why I’m convinced most enterprises will in the future use some form of hybrid infrastructure. Mission critical data and applications will run on mission critical infrastructure. Less critical applications or ones that need to be as close to endpoints as possible will run on cloud infrastructure or a distributed network of colocation facilities. So yes, we are headed back to the on-premise future, but it is a hybrid one in which the cloud plays the starring role.