Introduction
In this article, I would like to discuss the economic aspects of the question “public cloud or own infrastructure”. Right up front: I am explicitly not talking about clearly delineated cloud services such as Microsoft 365, but about the hosting of virtual machines and purchased applications such as ERP systems. In recent years, public cloud providers such as Microsoft Azure and AWS have gained enormous popularity because they offer companies flexibility, scalability and cost efficiency. Despite these advantages, running your own infrastructure often makes sense. Especially for continuous workloads, as is common in many enterprise applications, the public cloud not infrequently offers strategic disadvantages at higher costs.
Costs
Although public cloud providers offer usage-based payment and cost efficiency, costs can rise quickly with continuous workloads. Operating your own servers in a data center can be more cost-effective in such cases, as the acquisition and operating costs can be spread over a longer period. Depreciating the hardware over a realistic usage period is often orders of magnitude cheaper than the usage fees of equivalent cloud capacities. Of course, costs for the data center, connectivity, energy consumption and ultimately personnel must also be taken into account. Considering the fact that even an IaaS operation can by no means be manless and you definitely need IT skills to operate the virtual servers hosted wherever and to maintain the applications, the overhead from operating your own platform is manageable. Modern hyperconverged platforms such as VMware vSAN, Azure Stack HCI or Nutanix require relatively little maintenance and offer a high degree of stability.
Of course, operating one’s own infrastructure entails a certain base of costs, and this factor cannot be explained away. On the other hand, expanding the environment on the basis of these base costs is then much more favorable, since a server rack, for example, can accommodate 40 height units in any case. The initially procured top-of-rack switch cluster also does not require any new investments when additional servers are added.
The bottom line is that the operating costs of a company’s own infrastructure in a data center are often significantly lower than the usage fees of IaaS services, even when considering the full costs. It should also be noted that premium functions such as geo-redundancy or the establishment of general high availability are not automatic, even in the cloud, and entail both additional costs and higher configuration effort. In any case, every detail must be considered for an honest comparison.
My sample calculation assumes the following costs for a dedicated infrastructure:
Cluster
2x Hypervisor: around 30.000€
- Dual CPU: Intel Xeon Gold 5315Y
- 512 GB RAM
- 8x 1.92 TB SSD
- 2x dual port 10GbE network card
- Management Card Premium
- 3 years premium support
2x 10GbE 12 port switch: ca. 3.500€
Licenses
- VMware Essentials Plus 3 years: €6,500
- StorMagic SvSAN 12TB incl. 3 years support: €4,000
- 2x Windows Server 2022 Datacenter 16 cores: 8.000€
Implementation and commissioning
approx. 6.000€
Initial costs: 58.000€
Repayment loan at 4.5% annual interest rate: 1,725€.
Data center (monthly)
- 2x 1/2 rack: 800€ (Two fire sections or locations on campus)
- Cross Connect LWL: 150€ (connection of both sections)
- Connection 10Gigabit: 200€ (connection to the Internet)
- 5TB traffic: 100€ (20€ / TB traffic – connection to the Internet)
- Approx. 200€ energy costs based on hardware (55ct / kWh )
- Maintenance costs: approx. 1.000€ / month
Total monthly cost for 36 months: 4.175€
We can see here that even with a conservative calculation basis, the total costs of an autonomous infrastructure are below the costs of a comparable cloud environment. Note that we are talking here about a redundant, fully licensed Microsoft environment based on VMware. If you now consider that you can also rent a quarter of a rack or obtain individual switch ports with their own VLANs for use in our Advanced Housing, there is still significant savings potential here. Furthermore, the scaling of the environment based on the created framework conditions can be done quite cost-effectively compared to the cloud. An authentic comparison calculation is always interesting in any case and sometimes leads to unexpected results – in both directions!
With this article I want to make you think. It is neither against the cloud nor unreservedly in favor of the company’s own infrastructure. On the one hand, there are still various strategic aspects, which I have discussed and will discuss in other articles, and there are also various intermediate solutions, such as our advanced housing with managed storage or server rental.
(Please don’t nail me down on any amount, I deliberately set rather higher costs!)