You have a powerful server, but an office environment does not do it justice: everything stops during a power outage, the room heats up in summer, and upload speed falls short. Colocation (server hosting) solves exactly this problem: the hardware is yours, while uninterrupted power, cooling, and network infrastructure belong to the data center.
In this guide we cover what colocation is, who it makes sense for, and how to read the U, amp, and bandwidth line items in a quote, step by step.
What is colocation?
Colocation means placing a physical server that belongs to you in a data center’s cabinet (rack) and running it there. The data center provides uninterrupted power (UPS and generator), precision cooling, fire suppression, physical security, and a high-capacity internet connection. The operating system, the applications, and the hardware itself remain entirely under your control.
The point most often confused is how it differs from a dedicated server. With a dedicated server, you rent the hardware from the provider; in the event of a failure, part replacement is the provider’s responsibility. With colocation, ownership and maintenance of the hardware stay with you, and you purchase only the hosting infrastructure from the provider.
Who does it make sense for?
It is a serious candidate for anyone who needs to go beyond standard configurations. The typical scenarios are these:
- Specialized hardware requirements: GPU cards, large disk pools, hardware security modules, or software whose license is tied to a physical machine are often not found in rental server catalogs.
- Existing investment: If you have already spent money on hardware, moving the machine to a data center instead of keeping it in risky office conditions protects that investment.
- Data control: For teams that have to know which disks their data sits on and handle disk destruction under their own procedures, ownership is critical.
- Long-term, steady workloads: For a system whose resource needs will not change for years, buying hardware once can cost less in total than paying rent every month.
On the other hand, if your workload is small or changes rapidly, buying hardware kills flexibility. For projects like these, virtual solutions such as a VDS server are both more economical and more practical. If you are undecided, the following table offers guidance:
| Situation | Suitable solution | Why |
|---|---|---|
| You have your own hardware and want full control | Colocation | The hardware investment is preserved and infrastructure worries end |
| You need a physical server but do not want a hardware investment | Dedicated | Hardware and part responsibility sits with the provider |
| Resource needs vary and the initial budget is limited | VDS / VPS | Scales quickly, with no hardware cost |
Reading the quote: U, amps, and bandwidth
Colocation quotes generally consist of three line items: cabinet space (U), power (amps), and bandwidth. When you compare, you need to read all three together; a plan that looks cheap on one item can turn out expensive on another. If your needs amount to only a few U, you rent space in a shared cabinet; teams that want full isolation and room to grow move to a private cabinet.
U: cabinet space
U (rack unit) is the unit of vertical space in a cabinet; 1U = 4.445 cm. A standard server takes up 1U or 2U; disk-heavy storage machines can go up to 4U. Because tower cases do not fit efficiently into a rack, they are usually subject to a higher space charge; consolidating the machine you will send for hosting into a rack-type chassis saves money over the long term.
Amps: power consumption
In most quotes, power is capped in amps. To learn your server’s actual consumption, look at measurements rather than the PSU label: having an 800W power supply installed does not mean the machine draws 800W. On servers with IPMI support, you can see instantaneous consumption with the following command:
ipmitool sensor list | grep -i -E "watt|pwr|current"The amp calculation is simple: watts ÷ volts. A server drawing an average of 300W on a 230V supply uses roughly 1.3 amps. If you exceed the limit, you pay an additional power charge; for that reason, base your figure on consumption under load rather than at idle.
Bandwidth and traffic
Three separate concepts are frequently confused here: port speed, commit, and traffic quota. Port speed is the upper limit of the physical connection (1 Gbps, for example). Commit is the guaranteed bandwidth you can use continuously. Billing is handled either on total monthly traffic (GB/TB) or by the 95th percentile method: traffic is measured in five-minute samples, the top 5 percent of readings is discarded, and the remaining peak value is billed. Short traffic bursts do not inflate the invoice under this method.
Think of location as part of bandwidth as well: if your users are mainly in Türkiye, a cabinet in Istanbul delivers noticeably lower latency than an equivalent one abroad. The same speed commitment feels much better when it is close to the user.
Total cost advantage: renting or owning?
The monthly fee for a rented server actually covers two things at once: depreciation of the hardware and the hosting infrastructure. With colocation, these two are separated. You buy the hardware once; your monthly payment consists only of U, power, and traffic, and it stays noticeably below the monthly cost of a rental server in the same class.
For a sound comparison, use a simple formula: (hardware cost ÷ planned months of use) + monthly colocation fee. Compare the resulting figure with the monthly fee of an equivalent rental server. If you are going to use the hardware for 4-5 years, the depreciation share shrinks and, in most scenarios, the balance tips toward ownership.
To make it concrete: if you paid D lira for a machine you plan to use for 36 months, your monthly depreciation is D ÷ 36. Add the hosting fee on top and set it side by side with the monthly price of a rental server with the same processor, RAM, and disk configuration. When a rental contract ends, you are left with nothing; with colocation, at the end of the term you still have working hardware that has resale value.
Factor in the hidden costs as well: you supply a failed disk yourself, spare part stock is your responsibility, and physical interventions (remote hands) are billed separately at many data centers. On critical systems, dual power supplies and RAID redundancy reduce these risks considerably.
Before you ship your server: a checklist
Once the machine is inside the data center, every physical intervention means time and, in most cases, a fee. Complete the following before delivery:
- Configure the remote management interface such as IPMI/iDRAC/iLO, change the default password, and confirm that you can reach the console over the network.
- Check disk health: do not send a disk on the journey if the
smartctl -a /dev/nvme0n1output shows reallocated sectors or a high wear warning. - Always include the rail kit; a server without rails cannot be mounted properly in a cabinet.
- Enable the “power on after AC loss” setting in the BIOS so the machine turns itself back on after a power outage.
When you choose where to host, look at the data center’s standard and network capacity. SunucuCenter’s colocation service is provided in the Equinix data center in Istanbul; with a total network capacity of 80 Gbit/s, a 99.98% uptime commitment, and 24/7 support, your hardware runs under conditions it would never find in an office environment.
Frequently asked questions
How do I deliver my server to the data center?
Hand delivery and insured shipping are the two most common methods. Before delivery, schedule an appointment with your provider for cabinet and power assignment; send the rail kit and any spare parts along with the server.
What happens if my hardware fails?
With colocation, sourcing parts is up to you; data center staff perform physical interventions such as disk replacement or a restart at your request. Thanks to the remote management card, you can also resolve many issues yourself without going to the data center.
When does a dedicated server make more sense than colocation?
If you do not want to make a hardware investment, if you are running a short-term project, or if you want to transfer the risk of part failure to the provider, a rented dedicated server is the more suitable choice. For long-term workloads that require specialized hardware, ownership is the advantage.
Leave a Reply