TranslateProject/sources/talk/20191017 Data center liquid-cooling to gain momentum.md
DarkSun 1ba9fceedb 选题: 20191017 Data center liquid-cooling to gain momentum
sources/talk/20191017 Data center liquid-cooling to gain momentum.md
2019-10-19 01:50:50 +08:00

6.3 KiB
Raw Blame History

Data center liquid-cooling to gain momentum

The serious number-crunching demands of AI, IoT and big data - and the heat they generate - may mean air cooling is on its way out. artisteer / Getty Images

Concern over escalating energy costs is among reasons liquid-cooling solutions could gain traction in the data center.

Schneider Electric, a major energy-management specialist, this month announced refreshed impetus to a collaboration conceived in 2014 with liquid-cooling specialist Iceotope. Now, technology solutions company Avnet has been brought into that collaboration.

[Get regularly scheduled insights by signing up for Network World newsletters.]

The three companies will develop chassis-level immersive liquid cooling for data centers, Schneider Electric says in a press release. Liquid-cooling systems submerge server components in a dielectric fluid as opposed to air-cooled systems which create ambient cooled air.

BrandPost Sponsored by HPE

Take the Intelligent Route with Consumption-Based Storage

Combine the agility and economics of HPE storage with HPE GreenLake and run your IT department with efficiency.

One reason for the shift: “Compute-intensive applications like AI and IoT are driving the need for better chip performance,” Kevin Brown, CTO and SVP of Innovation, Secure Power, Schneider Electric, is quoted as saying.

“Liquid Cooling [is] more efficient and less costly for power-dense applications,” the company explains. Thats in part because the use of Graphical Processing Units (GPUs) is replacing some traditional processing, and is gaining ground. GPUs are better suited to data-mining-type applications than traditional processors. They parallel-process and are now used extensively in artificial intelligence compute environments and processor-hungry analytics churning big data.

“This makes traditional data-center air-cooled architectures impractical, or costly and less efficient than liquid-cooled approaches.” Reasons liquid-cooling may become a new go-to cooling solution is also related to “space constraints, water usage restrictions and harsh IT environments,” Schneider said in a white paper earlier this year:

As chip density increases, and the resulting rack-space that is required to hold the gear decreases, the need for traditional air-based cooling-equipment space keeps going up. So even as greater computing density decreases the space the equipment occupies, the space required for air-cooling it increases. The heat created is so great with GPUs that it stops being practical to air-cool.

Additionally, as edge data centers become more important theres an advantage to using IT that can be placed anywhere. “As the demand for IT deployments in urban areas, high rise buildings, and at the Edge increase, the need for placement in constrained locations will increase,” the paper says. In such scenarios, not requiring space for hot and cold aisles would be an advantage.

Liquid cooling would allow for silent operation, too; there arent any fans and pumps making disruptive noise.

Liquid cooling would also address restrictions on water useage that can affect the ability to use evaporative cooling and cooling towers to carry off heat generated by data centers. Direct-to-chip liquid-cooling systems of the kind the three companies want to concentrate their efforts on narrowly target the cooling at the server, not at the building level.

In harsh environments such as factories and industrial IoT deployments, heat and air quality can hinder air-cooling systems. Liquid-cooling systems can be self-contained in sealed units, thus being protected from dust, for example.

Interestingly, as serious computer gamers will know, liquid cooling isnt a new technology, Wendy Torell points out in a Schneider blog post pitching the technology. “Its been around for decades and has historically focused on mainframes, high-performance computing (HPC), and gaming applications,” she explains. “Demand for IoT, artificial intelligence, machine learning, big data analytics, and edge applications is once again bringing it into the limelight.”

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.


via: https://www.networkworld.com/article/3446027/data-center-liquid-cooling-to-gain-momentum.html

作者:Patrick Nelson 选题:lujun9972 译者:译者ID 校对:校对者ID

本文由 LCTT 原创编译,Linux中国 荣誉推出