There are fewer and fewer reasons to run your own data center. But some of them are very convincing. […]
In its simplicity and convenience, the cloud is displacing the desire of many users to have their own data center. This is not least due to the speed of innovation in the cloud: The move away from the data center is also due to the fact that cloud service providers are always convincing with innovative products that save time and money.
But with all the undeniably good reasons for switching to the cloud, there are also good arguments for resisting the trend and keeping your own data center running. Here are nine reasons to run at least part of your code on-site in a rack owned by your company.
The cloud offers many advantages for globally active companies – for example, to support remote workers. However, if your employees are on the same company premises and access the same servers, there are definitely disadvantages if the data must first circumnavigate the globe in order to be processed in the cloud. Local servers are faster. In addition, fewer network jumps mean fewer sources of error. If the bits never leave the building, you can get by with less bandwidth. These are good reasons to operate servers close to the scene of the incident.
Undoubtedly, the cloud can have a liberating effect, because the providers take away a lot of the hassle that the operation of servers, the purchase of machines or the installation and maintenance of software bring to their customers. But it can also be reassuring to have these tasks in your own hands. Especially if you want to operate your own solutions, the friction losses during a cloud migration are too high. The time savings that switching to the cloud can mean are not always worth the effort in such cases.
An example: In one of my projects legacy code was used, which required an old version of Python. However, the cloud provider used the latest version of Ubuntu, which, in turn, required a newer version of Python. I could decide whether I wanted to either mess around with different versions or install my favorite version of Python on my own computer. The latter was much easier than rewriting the code.
Cloud companies need to make all customers happy. Many different applications use the same services in a multi-tenancy world. To commit to a cloud service means to have to get along with the neighbors. In extreme cases, however, they can be malignant. Attack techniques such as Rowhammer have shown that it is possible to compromise other users on the same hardware. Of course, this is not (yet) a widespread problem.
Nevertheless, one of the great advantages of having your own data center is not having to bother with the neighbors.
Modern contracts are not set in stone, and most often they are no longer even written on paper. If problems arise, it may happen that providers simply dismiss their customers with the indication of a violation of any undefined clauses in the terms and conditions. The Internet forums are overflowing with sad stories of developers and companies who ended up getting a termination from their cloud provider. In some cases, the provider companies even do without it and simply turn on the cloud tap.
Maybe you have good lawyers. Perhaps you also think that such stories are exaggerated and do not happen to you. On the other hand, the probability that cloud providers admit mistakes and voluntarily renounce revenue seems low. There is no doubt that the number of legal pitfalls decreases when companies retain control over their hardware.
Many cloud providers have to defend themselves from the accusation that their services are incomplete. Some even deliberately do not specify telephone contact options, others generally do not respond to e-mails. On the rare great stories of employees of a cloud company who have put in a lot of effort for their customers, at least ten rants about nameless and faceless data octopuses appear on the Internet forums.
Managers will quickly receive answers from those responsible in their own data center. In sitcoms, jokes are often made about how rarely you get to see the internal IT support, but there is no doubt that the colleagues would like to continue to receive their salary.
The latest hardware is also always the most expensive. If your workloads are forcing you to deal with large fluctuations, it might really make sense to rely on the cloud. However, if the tasks to be mastered are rather unexciting and predictable, you can save a lot of money if you are satisfied with servers that have already had a few years under their belt.
Of course, there are also hidden costs to consider, for example, older devices break down more often. However, if you can cope with occasional downtime and also have employees who are able to repair the hardware, it is a cheap alternative to use hardware for a longer time.
Companies with highly fluctuating but generally predictable computing loads usually cope best with the cloud. Video streaming services, for example, reliably experience a peak on Friday and Saturday evenings. Therefore, use the cloud computing power for a few hours and go back as soon as the users go to sleep.
If your company continuously requires high computing power, running your own data center can be cheaper. Paying for a cloud machine 24 hours a day, seven days a week is expensive – even after deducting discounts. In addition, the budget calculation for a competitive local data center is easy if you plan with a 24-hour operation.
The pandemic has turned the world of commercial real estate upside down. Some companies are sitting on unused premises, for which the lease expires only in a few years. In such a case, it may be worth running a few server racks there instead of going to the cloud.
Some companies want to make do with as few staff as possible in order to reduce their costs. Running your own data center with the appropriate employees can be expensive. These are the costs that are the hardest to justify for a CIO. The balance will be better if the employees in the data center also perform some other tasks. You will not get such services from a cloud provider. And he will not tell good jokes at the coffee machine either.
*Peter Wayner writes, among other things, for our US sister publication InfoWorld.com and is the author of various books – including on the topics of open source software, autonomous driving and digital transactions.