John Weale’s past work has spanned the critical facility field. It includes the design of data centers, laboratories, semiconductor fabrication clean-rooms and assorted high-performance "green" school and office buildings. His data center work has included measurement and benchmarking of numerous existing facilities, design of a university supercomputer center, data center efficiency audits, as well as the design of multiple data centers that use free cooling and economisation.

DatacenterDynamics FOCUS: What, in your opinion, have been the biggest recent advancements in cooling technologies and where do you think those technologies are, or should, be going into the future?

John Weale: The biggest advancement in cooling technology for computers is the ability to start cutting it out completely – to simply reduce the amount of cooling equipment that is actually needed to keep these servers up and running.

Where I see we need to go and where the market is slowly heading is in manufacturing servers that can accept higher air temperature for cooling. We would eventually get to the point where complex mechanical systems to keep the servers operating are just not needed.

We’re not talking about a huge jump here, either. We’re talking about taking servers that can already operate normally in 72-degree Fahrenheit temperatures.
Everyone pretty much accepts they are not going to experience meltdown at 90-degree air intake. If you just came up another 10 degrees, you could eliminate the massive cooling systems that are currently in use. This does, however, require a shift in people’s perceptions. They have to be able to accept that servers are not fragile devices that are will break.

That’s where I see things really going. I know the market is starting to demand that from servers – it’s simply another design factor … that server manufacturers can accommodate. They just haven’t been asked to before. Ultimately, cooling will become less important and the biggest advance in cooling devices is going to be that they simply will not be needed.

DCDF: How was this common perception of servers as fragile devices formed?

JW: There is the legacy of how data centers evolved. Tight humidity standards are extremely important when you have a punch-card deck. Obviously, we don’t have punch cards anymore.

There are some specific data center devices where humidity is more important than others. If you have a lot of tape drives or something of that nature, it is critical. The devices and the equipment have advanced beyond that in many cases and they are continuing their forward trajectory. As customers care less about humidity, they are now demanding that servers accept higher temperature air and manufacturers are starting to provide that.

In the past there were devices that would fail and sometimes fail in a catastrophic manner. So, there is a great deal of "this is the way we have always done it and 20 years ago, a company went bankrupt because it lost power and its servers melted down and it lost all its business for three days and never recovered".

In the past there were myths that said you had to have these high tolerances and it was never questioned. But people are starting to question them now. They go to server manufacturers and ask: "Why does my daughter have a terabyte music server under her bed that she piles towels on top of and it’s been fine for three years, while I’m buying a $10,000 rack and you say it can’t take temperatures of 78 degrees for more than three hours?"

More often, though, customers assume the servers are fragile – they are not. But the ones that are will start to lose business, ultimately, because it’s too expensive to produce these tightly controlled spaces if you have a competitor that can produce a computer that doesn’t need it.

DCDF: How focused are leading server manufacturers on that front?

JW: Some manufacturers are beginning to provide more efficient servers. They are taking advantage of the marketing side of being able to sell a "green" server. There is an increased visibility of the cost of energy in these facilities and some data centers are starting to become limited.

They have 10mw from the utility but getting another 10mw is cost-prohibitive, making them determined to get as much computing power as they can for those megawatts. This plugs right into the idea of creating more efficient machines. I think server manufacturers are starting to hear that from the clients and it is now a priority. But servers are still a commodity item and so it is cost first, then power.

Looking forward, the ability to have a server tolerant to higher temperatures will become more important until it is just assumed that instead of keeping your server cool to the tune of 72-degree, you now must keep the server below 100 degrees, just as you would elevator machinery. The only issue being that the server will need to be made larger to get a little more air through.

You may also have to use liquid cooling, which is already provided for the densest supercomputers, or high-performance computers. It is just like a fluid loop that you might use in a car. If you are beginning to see energy densities equivalent to a car’s waste heat, you will need to provide a cooling system equivalent to that. That’s where I see things heading in the future. Whether or not we get there, will remain to be seen. If energy remains relatively cheap – as it is now, for example – there may not be the market forces to push it there. But as energy becomes more expensive, a commodity market must start to consider alternatives, and this may be one alternative worth looking into.