Capital Expenditure versus Operational Expenditure

Header-NL-Ref-1-Jun-13-2023-06-38-30-2350-PM

 

Technology is constantly evolving. Every three to five years, the best practices that have just been accepted as standard change again. A retread of the ICT infrastructure is appropriate every ten years at the latest, so as not to fall so far behind the state of the art that you have to struggle with compatibility problems. Research and comparisons are made in order to select the optimal setup for one's own needs. After the euphoria about the new possibilities of the technologies and their capabilities, however, comes quickly the disillusionment about the price. The latest and greatest has its price and the manufacturers are more than aware of this.

The problem with the usual planning of network and server landscape is that you have to plan so that the theoretically required peak load can be covered by the new equipment. This often leads to the selection of configurations that go far beyond everyday use. The resulting costs are rarely or never justifiable in times of shrinking budgets. However, there is an alternative to this approach.

More and more companies are no longer willing to invest significant amounts of capital just to buy something whose real value is, if not questioned, at least often doubted. In the private sector, the rental, or as-a-service, model is already available for almost everything. Mobile phones are included in the contract, Netflix & Co. deliver movies and series on demand, and even cars can be obtained from Volvo as a subscription. This expectation has also shifted more and more into B2B in the last two years. Many services can already be purchased as a consumption model. Now the large manufacturers of ICT equipment have also started to include this model in their portfolios.

Due to the emergence of outsourced capacities in large data centers, it is now possible to subscribe and unsubscribe to many services on a monthly basis according to performance. Servers are only virtual, databases are completely off-site and even the management of the company network is controlled via virtual machines. At the individual locations, only switches and end devices such as shuttle PCs, printers and telephones are necessary. Even expensive dedicated lines are unnecessary thanks to increasingly reliable consumer-grade Internet.

But what if you are not willing to entrust your data to others and hope for their discretion. Even if the reputation of providers like AWS or Azure is indeed very good, one must not forget that they primarily belong to corporations that primarily earn their money with data. As a public carrier, the disclosure of data is generally not up for debate and thus cloud offers fall completely out of the possibility pool.

However, in order to take advantage of the flexible pricing of cloud offerings while still operating the hardware on premise, manufacturers have come up with the Infrastructure as a Service model.

With this variant, the devices are rented directly from the manufacturer and thus receive the advantages of cloud solutions such as performance-adjusted pricing, transfer of functional responsibility, scalability and the avoidance of capital expenditure for acquisition. However, the data stored on the hardware is on premises and at no point exposed to security concerns in the cloud. Thus, user data and company secrets can be stored securely.

Of course, this raises the question of how scaling should take place when the devices are on site. The manufacturer delivers pre-configured devices that correspond to the expected normal consumption and takes care of their maintenance, updates and possible replacement. If more power is needed at short notice, it is added via cloud management for the duration of the workload. In this way, only the costs that are actually used are due per month. The data does not go into the cloud but remains on premises or in the data center.

In this way, the normally very high capital costs for the acquisition of hardware can be allocated to the operating expenses and, in parallel, exactly the desired performance can be obtained.