Basic introduction
For information service enterprises (such as ISP / ICP / ISV / IDC), first consider physical parameters such as the volume, power consumption, heat generation of the server when selecting the server. Because information service companies typically use large-scale dedicated computer rooms to deploy and manage large number of server resources, computer rooms typically have strict security measures, a good cooling system, multiple backup power supply systems, and their computer room cost is quite expensive. How to deploy more servers directly to a limited space directly to the service cost of the company, usually use a mechanical size that meets a rack server of 19-inch industrial standards. There are also a variety of specifications, such as 1U (4.445 cm high), 2U, 4U, 6U, 8U, etc. Usually 1U's rack server is most space, but the performance and scalability are poor, suitable for some business relative to fixed use. The products with more than 4U are high, good scalability, generally support more than 4 high-performance processors and a large number of standard thermal plug components. Management is also very convenient, manufacturers usually provide people's corresponding management and monitoring tools, suitable for large visits, but large volume, and space utilization is not high.
The advantages of the rack server, compared to other two style servers. Rack servers save space relative to the tower server, but heat dissipation is slightly worse because there is a compact space.
Clever problem
What is the fever?
is different from a server for each workload, a server-like server, and a virtualization manager can support 10, 20 or more workloads. The frame space of the facility may be idle due to various loads after virtualization.
At the same time, the chip is created by a higher density transistor stage manufacturing process and a lower clock speed. Therefore, the spiral rise of the processor core number will hardly affect the energy consumption of the rack at least when the device is updated.
Reduce the scale, there have been more fully utilized servers in the data center, so the racks need to be reduced, which has changed how we applied cooling. Unlike the cooling of the entire data center, the air convection in space is used to achieve space in space, such as the thermal / cold air channel to achieve the inhibition strategy, and reduce the operating area in several smaller spaces, even in some racks. These fever are processed by internal cooling systems within the line or the rack, and even turned off the equipment room air conditioner (CRAC).
In addition, the United States Heating, Refrigeration and Air Conditioning Engineer (ASHRAE) organization also suggested to enhance the effective server air inlet temperature to 27 or even 32 degrees Celsius.
With the development of these energy management, there is a case where hot and cooling is inadequate, and it is generally caused by improper design of improper or bad.
Hotspots and other cooling problems
Even with the best suppression strategy and high efficiency cooling system, the server hotspots in the rack will still be generated due to calculation equipment .
Accidental obstacles or air flow paths can cause heat. For example, remove the guard of the server rack, let the air flow into the rack plan, will weaken the air flowing to other servers and increase the exit temperature.
greatly increases server energy consumption, which will also cause heat dissipation. For example, replacing several 1U servers with a high-grade blade server system, which greatly enhances the energy overhead of the rack, and the insufficient air flow will directly affect all module components of the blade machine. If the cooling system is not designed for such a server, it is likely to have a hot spot.
When increasing the servant rack density, operational organizations need to consider investment data center infrastructure management and other system management tools, collect data from rack internal heat sensors and generate reports. They can find more than fever limitations and take the necessary measures, such as notifying technicians, automatically calling workload migration or shutdown systems to prevent facilities from failing to fail.
When the server rack plangens generates hotspots, the IT team can reassign hardware. Unluckled with a single rack, if space is allowed, half or one, two devices to other racks, or turn off overheated systems.
If space is not enough to redesign, add some movable, self-contained air conditioning, and can be used in the data center. If the rack uses a compact line or the cooling unit in the rack, set the temperature point to achieve a cooling effect than the cooling device is more efficient than opening the closed cell.
Long-term mitigation strategy
From the long run, breakthrough technology helps heat management.
The water-cooled rack can transmit cooling water through the door or other path. Water-cooled racks can solve most of the heat-free problems - especially when there is only only low temperature air and high temperature air to dissipate heat.
China immersion cooled technology can immerse the server in a bathtub filled with mineral oil, but non-conductive, non-corrosive cooling substances. This technology is expected to achieve high efficiency, with almost no noise and heat transfer close to zero loss.
However, these hot technical options are more suitable for new data center architecture, not ordinary technical cycle updates.