09 September 2020

Efim Mirochnik

VP of Data Center Operations in Acronis

What is more profitable: colocation or own data center?

There are many nuances in choosing between building your own data center and renting a place from a colocation provider: this includes the size of the company, the life cycle of the equipment, and the financial indicators by which the company is assessed. Efim Mirochnik - VP of Data Center Operations in Acronis, a special guest of our blog, has shared his views on how to choose the best option.


One should always consider clouds while defining IT strategy. But there are scenarios when the cloud technolog is not applicable. For instance, your service relies on specific server hardware (like mainframe for a bank or ASIC for crypto mining). Or you don't trust clouds. Or they are too expensive for your tasks. Then there are only two classic choices left.

An own data center is premises purposed for server hardware placement. A data center provides physical security of hardware, electric power, cooling, and network connectivity with the expected reliability level. The data center owner is responsible for its architecture and service quality.

Colocation is a service of data center space rental. The service has an agreed cost and reliability level with a money guarantee. The server hardware of a colocation client is placed on the provider’s premises. Colocation may provide additional services with smart-hands as one of the most important service components. Smart-hands are technicians in the data center that may perform maintenance with the server hardware.

There is no universal answer to the selection between colocation and an own data center. Various factors influence the decision:

  • company size and its goals (including financial and strategic ones)
  • risks and spending evaluation
  • local country and region specifics.

Company valuation by EBITDA

The EBITDA (earnings before interests, taxes, depreciation, and amortization) defines some companies' market value. Thus, EBITDA is often used to evaluate startups. This indicator includes only operational expenses. Expenses of building an own data center consist mostly of capital expenses. They are accounted for in the depreciation and amortization line, which is outside of EBITDA. It improves the visible financial valuation of the company.

Equipment lifecycle

Own data center requires complex planning. The reason is that the lifecycle of different data center components is different. Thus, cooling equipment may have 6-7 years, UPS has ten years, while UPS batteries may require replacement in 3-4 or 10 years depending on their type. Ideally, the data center should be designed to minimize unexpected further investments.

IT growth speed

The next important factor for planning is the growth speed of the IT technology. The requirements to them now and in 4 years will be different. The design and construction of the data center may take 6-18 months. Therefore, it's better to plan with a long term perspective. If IT infrastructure grows twofold per year, it is advisable to introduce modularity or equipment commissioning phases. They allow to increase data center capacity according to demand growth and not to spend investments onto initially unused capacity

Equipment maintenance cost

Both data center and server hardware require maintenance for the owned data center. Such maintenance needs qualified technical personnel: electricians, cooling engineers, server hardware technicians, physical security, health and safety specialists, and round-the-clock monitoring team.

A specialized organization can assume the task of data center maintenance. An owned data center with contracted maintenance will be an intermediate option between colocation and full ownership.
The farther the data center is from the places with qualified personnel, the more expensive and difficult its maintenance costs. But the land costs will probably be cheaper.

Colocation removes most of these complexities and hands them over from the client to the colocation provider. The colocation service requires regular payments, which are operational expenses and reduce EBITDA. On the other hand, no significant investments are needed. One can start using colocation much faster compared to building a data center.

Colocation costs consist of several parts:

  • Place in the rack or several racks of the required sizing. The rack unit is a standard measurement for a place in server racks. Full racks can be installed in a shared room, in a cage, or a suite.
  • Electricity is often the most expensive part of the colocation costs:
    • Maximum electric power (kiloWatts, kW) is the limitation of power that client hardware can consume. Thus, the 6kW limit per rack means that the server and network equipment's total consumption should not exceed 6kW at any given moment in time. Usually, this parameter defines monthly fixed rental payments per rack.
    • Electric power consumption per month (kiloWatt*hours, kWh) measures used power. This power consumption value is multiplied by the power cost and power usage efficiency (PUE) coefficient to calculate the consumed power cost. The PUE coefficient depends on the operational characteristics of the data center. Sometimes actual power consumption may not be separated or explicitly metered. Then it is included in the provider's service bundle cost to make monthly payments fixed.
  • Network connectivity is local connectivity inside of the data center, access to ISPs and the Internet.
  • Contract terms are contract period, ability to terminate the contract early, notice period, and automatic prolongation terms. Contracts in many countries may last for many years without the ability to cancel the contract before its expiration.
  • The escalation coefficient is a yearly increase percentage for fixed payments. Usually it corresponds to the inflation rate.
  • Smart-hands, monitoring of electricity quality and continuity, consumed power, current environment temperature and humidity in various points, monthly reports are additional services provided by the vendor.

Colocation costs comprise of two parts:

  • Non-recurring charges (NRC) cover expenses of infrastructure preparation for server hardware
  • Monthly recurring charges (MRC)

The overall conclusion requires to calculate the data center expenses and get proposals from colocation providers.

It is convenient to use the Total cost of ownership (TCO) for comparison. This will require the following:

  • define the period which we want to compare
  • colect all planned capital (CAPEX) and operational (OPEX) expenses for each scenario within the period
  • compare the results

TCO = CAPEX + OPEX

The most profitable option will be evident.

Not so obvious complexity appears when calculating planned expenses. It requires a reasonable understanding of the market and strong price level knowledge. If a specification is available, either in-house technical specialists or a specialized design company can make high-level planning. However, one is expected to pay for their service as well.

It is possible to step away from the "what's more profitable" task and switch over to "what's more reasonable" question. Such an approach allows us to throw out an unsuitable option. It will be sufficient to review company size and its tasks.

A small company with an office

Constant saving is the usual situation in small companies. If the company already pays for the office, it is possible to dedicate a small pantry or a room for the rack. One can install a few servers there and call it a data center. Additional cooling will probably be required. But most likely, it will be good enough even without cooling.

Such data center expenses are minimal: a dedicated room or a communication cabinet, additional rack UPS, and air conditioner.

Such a solution allows to set up a cheap and fast network because all the workplaces and servers are in the same location. Another benefit is the ability to avoid a long term contract for colocation.

Disadvantages of such an owned data center compared to colocation are numerous:

  • non-reliable
    • Electricity in the office can be turned off accidentally or regularly for maintenance.
    • Neighbors add risks to harm the equipment: flooding, breaking walls, or damaging cabling.
    • Central cooling in the big office may be turned off in non-business time. It may lead to overheating at the weekends. It may be exceptionally inconvenient to find out about this in the period of financial statements submission.
  • Equipment in the office may be too noisy or disturb nearby employees.
  • This solution is difficult to scale.

Small companies might prefer to choose colocation in the following cases:

  • If the company requires a reliable solution that an office cannot guarantee.
  • If the internal infrastructure is required to provide round-the-clock access. It is especially vital during the COVID-19 situation when physical entry into the office can be closed or limited.
  • Office cannot provide the required level of physical security.

Medium company

The reliability requirements for IT infrastructure in medium companies are usually quite significant. The company might have several offices. Or clients might require high service quality. It means that IT infrastructure should be quite reliable.

If there are no own data center specialists, a better option would be to start with colocation.

There could be no colocation providers in the region. Then the company needs to decide on building its own data center.

Large company

Here are the essential factors influencing data center spending:

  • network
  • electricity
  • environment (climate, cooling agent: air or water)
  • building and land for data center placement
  • risks of human mistakes, political, technogenic, or natural factors
  • personnel
  • service audience

These factors can change their significance depending on the service. Thus, crypto mining may be successful and profitable in cold and places with cheap and reliable electricity, for example, in Siberia.

But this location will not fit mass interactive services if their primary audience is many thousands kilometers away. Long-distance cable laying is quite expensive. Also, network latency will increase, slowing the service down.

When one considers significant risks for big projects, it is easier to use an existing colocation service. However, it might not be more profitable.

If colocation is not available in the region, one might make a long-term agreement with a colocation provider. Your company may get a professionally built and run data center. The provider will have a guaranteed client who starts paying right after the data center commissioning. Such colaboration increases interdependency between the companies but reduces risks for both of them.

The larger the company is, the more significant its tasks can be. At the same time, big companies may have small and medium local objectives.

Thus general recommendations for selection between an own data center and colocation would be the following:

  • Own data center fitssmall companies or specialized tasks
  • Medium companies and standard installations will be more profitable in colocations.
  • Big companies consolidating infrastructure or having specific tasks more likely will have their data center more profitable. On the other hand, colocation is a preferred choice when IT is not a core competence,and colocation service is available.

Certain aspects may have not been taken into consideration in this recommendation, but it can replace complicated calculations.

Certain aspects may have not been taken into consideration in this recommendation, but it can replace complicated calculations.

ISPsystem recommends the DCImanager platform to manage equipment in your data center or at a colocation provider.

DCImanager helps to keep the infrastructure under control and simplifies its maintenance in a number of ways:

  • Manages servers, switches, networks and PDUs remotely;
  • Keeps inventory of equipment in data centers with accuracy down to one unit;
  • Indicates the equipment status and alerts about potential problems;
  • Automates installation of the OS and other routine tasks.

Efim Mirochnik

VP of Data Center Operations in Acronis