Types of data center design

IT professionals always have to learn, apply new ways to design the structure for data centers to get efficiency, ensure large capacity and easy to expand without any trouble (efficiency, capacity and scalability).

Those who see the usual way - using network cabinets and placing servers - are inefficient and should consider switching to pre-designed network cabinets, container and container groups (engineered rack, pod and container).

Over the years, data center managers often use solutions that simply increase the number of parts of the server (component server) when high-performance operations are needed, such as buying 19-inch ceiling cabinets and inventory. On that server 1-4U. This is a simple and easy to understand way when workload increases. But this way is no longer as simple as before.

Limitations of ceiling cabinets

When vendors increase the amount of power consumed on the device, placing multiple servers in the cabinet can lead to many unexpected problems. For example, if a high-speed CPU can run very hot, the data center will have to take proper cooling measures.

Using a network cabinet without thinking about how to place the electrical system or server parts can lead to hot spots and it is difficult to cool, thereby causing the device to quickly fail.

As virtualization is used more and more, technology professionals do not use separate server, network and storage systems anymore. Instead, they put them close together, usually in the same cabinet or cabinet to get the best performance.

Because of this, cooling is difficult. Database architecture management tools (DCIM - Data Center Infrastructure Management) like nlyte or Trellis need to run the 'what if?' Script. (if so) and determine the future state by fluid dynamics (CFD - Complex Fluid Dynamics) to see which points are hot.

A new approach in designing data centers can help overcome this problem. More and more suppliers are offering pre-designed network cabinets with rows or pairs as a module or creating a completely separate 'data center' with standard containers to store everything. necessary to run a workload.

The reasons for Data Center crash

Building modular data center / container

The system is structured as a module or the other is a pre-designed pod. Cisco is a pioneer with UCS (Unified Computing System), followed by a collaboration between VMware, Cisco and EMC with VCE vBlock architecture. Since then, many firms have also launched similar products.

Modular data center

These modular data centers are a complete, separate solution with system management, virtualization, storage, network, power distribution and cooling tools. With a single-row system, the cooler will be on the line, while in the double-row system, they will be placed in the heating / cooling corridors.

But a big problem with the modular system is that when it comes to expansion it will take a big step to add other modules or add small systems like pre-designed network cabinets. Either way will not be easy.

Finally, a modular data center / container is born. Initially, the suppliers believed that this was a specialized system, used in separate cases, such as when small data centers were used for a short time, or long-term use but were not qualified for facilities. .

Container systems can be put into one place and as long as there is electricity to use (sometimes water for cooling), the container can still operate. Finally, it was realized that the container system could be used in data center design.

Built-in network cabinets are suitable for small changes, while modular models are more flexible but still need to be built in place. Containers only need to be taken to the car to go to the place, unloaded, plugged in and used.

Container center

Companies like Microsoft combine modules and containers in their latest data centers. Here, containers are used for basic tasks and need to run quickly and use modules when needed more flexible.

Upgrade modules and containers

The biggest problem of containerized systems is too rigid. If you want to change, you will have to remove it completely and rebuild it. Containers are fixed in size so the equipment used is also very specialized. The data center manager will see replacing the container with the new one rather than fixing the old one.

Some suppliers recognize this problem and have a solution, such as leasing and when the time expires, replace it with a new equivalent system, take the old one for recycling. Intel also developed a closed high-temperature container, increasing the likelihood of equipment failure due to overheating but running more efficiently with minimal cooling. The container runs for a while, when the equipment is broken, it is allowed to over-engineer to run at the original speed, then replace it with a new container, the old one is used for recycling.

Dedicated network cabinets, modules and containers all play a separate role in the data center's future. The life span of self-designed network cabinet systems is also getting shorter. Data center managers should consider using a variety of built-in modular systems, planning how to cool and capacity issues for the system properly.

See more:

1. Learn about Amazon EC2 service
2. Implement these tasks first when transferring data to Windows Server 2019
3. Learn about RAID