Your top 5 liquid cooling quandaries answered, according to Omdia With generation after generation of chips and systems growing ever hotter, and datacenters increasingly under pressure to reduce their impact on the climate, liquid and immersion cooling technologies have steadily gained traction. Chipmakers and OEMs alike are investing heavily in liquid cooling. Intel, for example, plans to …
CHP (Combined heat and power) stations are amongst the most efficient ways of using resources. I did some early design work on them for a degree course in the late 1980's.
There is no reason not to use the energy left in the fluid to do something useful, and increase the overall efficiency and probably drive down operating costs of the datacenter. Stirling engines (https://en.wikipedia.org/wiki/Stirling_engine) are fairly easy to build, just get them driving a generator, and you get a portion of your used power back.
"What's more, unlike direct liquid cooling, immersion cooling is largely vendor agnostic."
Ok, but if it is vendor agnostic what about this:
""Not all companies are producing servers that are guaranteed to operate in immersion." This is a problem because the dielectric liquid used in these systems is relatively new and may breakdown or degrade certain components if they weren't designed to operate immersed in an oil or two-phase refrigerant."
Clearly there are problems.
Immersion cooling is extremely messy. You aren't going to be able to get support on any common server once you've immersed it, as it will be full of goop. Additionally, all fans must be removed from an immersed server, and fan monitoring disabled. You can't do this with all equipment. Additionally, servers need to be mounted with hot side up, so convection can circulate the liquid through the server. This requires special hanging racks.
And don't try using spinning disks in immersion cooling. The liquid will probably penetrate the bearing, seals, or drive motor. Anything with an electric motor in it, can't be immersed.
I loathe immersion cooling. The new non-toxic liquids perform well, but are hard to clean up. On one immersion trial that I was involved with, the circulation pump was leaking, and it was dripping this liquid. It created a huge mess. Plus, when equipment had to be pulled up from the liquid, you'd be covered with it. The particular liquid I was working with did not dissolve in most common detergents. The floor was slippery and we couldn't ever get it clean. You have to assume that anything you dipped into the immersion pool (ours was a long trough), was something you'd never want to lay an ungloved hand on again.
I can see a server having one tank for the dip-in stuff, and one dry rack for the storage...
... but now you have all the infrastructure you had for air cooling, plus the dieletric, plus means to connect the storage to the sunk racks that once were together.
Indeed, some folks should not have this at all.
And there is a joke about sunk cost lost in there somewhere.
On the HDDs/SSDs question.
We have not been brave enough to run live on HDD yet but it is being tested (soak test in every sense of the word).
Our live servers in liquid are SSD so not challenges in terms of operation, swap out is always going to involve taking a node off line, in our tank the SSD's are located at the bottom of the tank in preference so we have the power and ethernet cabling at the top, effectively reverse mounted servers. As long as you design around the fact you will not be able to hot swap drives.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
