Batteries and Data Centers

Estimated reading time: 34 minutes

So, grid stabilization is the real reason, the number one reason, why you would see a huge plant of batteries somewhere. And then, the second reason is because of the customer downstream. I mean, there is an upstream benefit, but there is also a downstream benefit. The downstream benefit is that a data center can now look to this battery for one, two, three, or four hours of runtime without the utility. Now, we’re hooking these things up to 120 kV, 230 kV, 345 kV transmission lines. We’re not hooking them up to a 480-volt service because we’re talking about 300-, 400-, 500-megawatt substations. So, we have to look at the voltages that these work at, and that’s not at a small village somewhere. I’ve worked on some data center campuses that, frankly, use more power than Baltimore, certainly more power than the town I grew up in. When you have that type of customer, they’re going to buy an awful lot of power. I mean, we’re talking about tens of millions of dollars a month in electrical charges.

And they want it to be refined power, which means they want it to be within what we call the ITI (CBEMA) Curve. You can look this up. It’s in IEEE standard 446, I think. CBEMA is the Computer Business Equipment Manufacturers Association, and ITI is the Information Technology Industry. They came up with a curve of time versus voltage deviation. You have to stay within this curve in order to satisfy the power requirements of most data centers and most computer servers.

Performance of these large battery plants normally doesn’t fall within that classification. They don’t satisfy that standard, but they can be made to do so. But you have to put in some extra engineering into that. That’s really where I come in because I can work with these large battery suppliers and say, “In order to make that into an actual UPS, you have to do this and that and the other thing. And then, if we can get that done, the data centers will be able to use this power as a backup in case that grid goes down.” Well, grids that are 120 kV and up are usually pretty reliable. We’re talking transmission grid. We’re talking national grid, right? They might go down for several hours, and that would be quite unusual, but not unheard of. If we are careful about where we site these things, we would be able to draw power out of two portions of the national grid that wouldn’t be ordinarily affected by the same event, whatever it might be; hurricane, flood, another derecho, another Sandy, or something like that.

So, those are the two main reasons to have a large plant. You’re going to be able to work with a utility to help stabilize the grid, and they’re going to pay you for the ability to get their hands on that power sink, which are flexible up and down. And at the other end, you’re going to have a customer who’s willing to pay you to have this power come to them refined enough to stay within the ITI (CBEMA) Curve classification.

Matties: When power drops out, is there any interruption in the service with the battery technology? Or is it an instant on?

Mosman: Normally, we’re going to have a quarter-of-a-second outage because most large battery plants that are being built solely for the purpose of utility grid stabilization, and they don’t have what we call critical customers downstream. What utility calls critical and what I call critical are two completely different things. If the utility loses power and gets it right back on in 15 minutes, they might not call that an outage.

But computer servers can only have power go away for one quarter of a cycle, not one quarter of a second. It takes one quarter of a second for you to open up a breaker, a physical mechanical breaker, especially one at a high voltage, in order to separate a battery plant and its load from the failing utility. You don’t want to have the battery feed into an upstream fault. All the inverters would just shut off and it wouldn’t serve any purpose, whatsoever.

You can, however, turn off an inverter almost instantly. But then, before you turn it back on, you have to open up a faulted utility input. Then you can turn the inverters back on and they won’t back-feed into a fault. Well, these breakers take a lifetime to open, from the viewpoint of a computer server. And so, in order for the battery plant to be a UPS, it must produce only have a one-quarter-of-a-cycle outage. It should be able to separate from the utility in, at most, not any longer than a quarter of a cycle. That’s maybe four milliseconds. Most breakers take 50, 60, 70 milliseconds to operate and another couple of tens of milliseconds in order for relays to work.

Johnson: So, you’re using the battery bank constantly and really using it as a way to make sure that you’ve got clean, steady power into the server bank. It’s working like an internet firewall.

Mosman: Well, it’s a UPS— an uninterruptible power supply. And uninterruptible is defined as a power supply that doesn’t allow any outages longer than a quarter of a cycle.

Feinberg: Right. In fact, what I’ve done here is I run uninterruptible power supplies coming off my solar backup. So, if that were to go down, I don’t want that quarter second or tenth of a second. I don’t want anything. I don’t want to even know it. So, these take over if that dies, they take over if the city dies so I never feel it. I’ve had it happen and I never even get a blip.

So, you’re using this battery power sometimes as the power, the moment-by-moment power, and sometimes backup. What’s the main thing you use to recharge these batteries? Is it diesel generators? Is it solar banks? Is it wind? Or is it a combination of all of them?

Mosman: Well, it depends on how you set it up, but more than likely, if we have engine generators it will be them. Let’s say we had an outage, and the outage were to last for several minutes or hours. The battery has to be recharged during that time. And so, it’s going to come from engine generators. If we don’t have any engine generators and we only have batteries to back our power up, then we have to wait for the utility to come back before it can recharge.

Feinberg: Right, right. That’s my point.

Mosman: But they turn out to be pretty expensive when you containerize engine generators. So now, what’s popular is building bespoke data centers. There’s a whole REIT industry built around that, real estate investment trusts. And there’s a lot of investment companies now that are starting to look at these types of buildings. They want to pour a whole lot of money into them. The trend is to get rid of engines, they don’t want diesel engine generators. I have specified, literally, thousands of large engine generators, two-megawatt to four-megawatt. But these things, they’re noisy. They’re dirty. They’re expensive, and take up a lot of space. We don’t want them anymore.

So, what’s happening right now is, we’re seeing if we can build a data center and back it up with batteries for one or two, or maybe three or four hours. And if we have an outage that’s longer than what our battery runtime is, we’re going to depend on geodiversity. So, the Big Five companies, and everybody else that wants to be a Big Five company, are building not one, but two or more data centers in different parts of the country. The advances in the transmission of information through fiber and networking have made it feasible to actually have geodiversity in data centers. So, they find it cheaper to build two data centers with no redundancy and maybe no engine generators because, if one goes down, it will just transfer all the traffic.

Matties: They just shift the load.

Mosman: You couldn’t do that in banks, brick-and-mortar companies. They only had one place. They had to contain it. It had to be secure. We now have Amazon. If you look at Amazon, how many data centers do they have? Fifty? Sixty? And they’re all connected.

Matties: And they do tout that when they’re selling the service, with that decentralized, multipoint program or center program, it’s probably the most robust and secure that you can have.

Mosman: Except that it still goes down. There was just in the news a story about how the Amazon guys lost some of their connectivity for things like Roku. All of a sudden, one day, Roku didn’t work for several hours until they could figure out what was going on. And what happened was, in a facility that they were leasing (from one of these real estate companies that make data centers for other people to use) have two types of servers. They have the headend servers. And then, they have the backend servers. The backend servers is where all the work is done. But all the switching goes on at the headend servers.

And so, the amount of traffic that you need to go through this data hall dictates how many of the headend servers that you’re going to have. And the way that they wire them up, each of these headend servers are connected to all of the other headend servers, so they can manage the traffic and disperse it equally to all of the backend servers that did all the work. Well, they noted a sudden step-up in load, thanks to COVID, and started adding more and more headend servers. But they forgot that the software that they were using had a limit as to how many connections they could make with other servers. Now, all of a sudden, their servers started hitting their maximum number of connection points, and they went off-line.

Johnson: At this point, Mike, I just want to get your take. Our readers, our manufacturers, have their facilities and they are running equipment in their own facility, usually significantly smaller than your data centers. But still, they’re running equipment in their own facilities that need uninterruptible power supplies. They might be using generators and the like. Is this a time for them to reevaluate what they may have as their myths or misconceptions and relook at battery technology as an uninterruptible power supply for their facility?

Mosman: Oh, absolutely. I think that the stabilization of the grid is becoming a very big thing. Do you think that we’re going to have less utility outages in the future? No. We’re going to have more as the system has been deregulated. The funding that utilities have had to repair and upgrade and maintain their facilities has been less, not more. When they sold off all these generating plans to private industry, they had incentive to not generate the VARs that are required to stabilize the grid because you get paid for watts but not for VARs, reactive volt-amps. So, they shut that off because it still costs money to generate them. But when you take the VARs out of the system, it gets all wobbly. That’s still going to become more and more a problem. And then, we’re putting in all of these windmills and photovoltaics, which are jiggling the system. The systems become more wobbly as more things are jiggling it. And that’s why they need the professor at Washington State University to figure out how they can build the grid and put in a control system that can manage all that. Well, there’s going to be a catch-up. The grid is going to have more outages. So, if you have a company, small or large, and you have a critical mission, that’s what you should do.

I design electrical systems for mission-critical facilities. No job too big, and no job too small. My three favorite words are “Follow the money.” My second three favorite words are “It’s all crap.” It’s true. All of this equipment that you buy, all these batteries, all this UPS, all these engine generators, it’s all crap. And because we’re trying to commoditize it, it’s starting to really become crap. My third three favorite words are “Trust your engineer.” Engineers, like myself, should be consulted by large and small companies.

I can remember when 100 megawatts was like, wow, how can we get 100 megawatts into a single building. Well, we’re building 100-megawatt data centers, regularly, but I think that we’re going to start looking at 500- to 1,000-megawatt campuses pretty soon. There’s going to be high pressure to keep the capitalization down. And so, we’re going to be looking at what is the cheapest crap we can buy because we’re “following the money.”

Well, that’s where people like me come in because we’re going to be the engineer that’s going to help the owners, who are chasing the money, keep from buying stuff that’s so crappy that it won’t work. The way that I have done this for almost 40 years now is, first of all, understand what the business mission is. What is this guy doing with his business? He’s following the money. He’s trying to earn money. He’s going to need his data center to do something that helps him earn money. And if it requires that that be up all the time, then we have to establish the criteria under which he operates his business. And once we have what the business mission is and what operating criteria is, then we’ll evaluate the scope of that. Is it going to be this big, that big? And then, we can discuss things like, “Here’s how much redundancy you have,” or... “Here’s what’s needed.”

It’s very complex because it gets into plumbing and security and fire protection and air handling and architecture and all of these things coming together. It’s not just, well, we’re going to throw a UPS down here and everything’s fine. It doesn’t work like that. That’s been my business. At my old company, CCG, I hired as many architects as I have engineers. So, if you’re talking about a small manufacturing company and you’re going to depend on power not going out, there’s probably 10 ways to skin the cat. And one of them is going to be better and costs less than all the others. That’s what you’ll want to know. What is the best? And what’s the cheapest?

Feinberg: We’re now making, I don’t know, hundreds of thousands or millions in batteries, what’s going on with the recycling of all these batteries? I mean, the number of batteries that are going to have to be recycled is growing exponentially. And not something you’ve heard about much lately yet, but with building them like this, the recycling is going to go like that.

Mosman: Oh, I’m the wrong person to be asking that question.

Matties: I would guess that’s an emerging market, though.

Mosman: I can’t see it not being an emerging market. For lead-acid batteries, everybody wanted that lead back. For newer batteries, it all depends on how much it’s going to cost to get the battery torn apart, get the elements out of it, and separate out all of the junk that polluted it. And that process, right now, is a little bit more expensive than going out and mining new metals and building new batteries. And so, really, it’s my opinion that recycling batteries must become part of the green revolution that data centers are undergoing everywhere. Right now, what a lot of these hyperscale data centers are very, very interested in, is the power that they are using dirty power or clean power? Is it sustainable? There’s a drive by 2025 to make these huge data centers totally sustainable, meaning that we’re going to use all wind, all solar, all nuclear, whatever power.

Nolan Johnson: Is there a shock on the supply chain currently, Mike?

Mosman: I don’t see any happening. I haven’t heard of any. There might be a few rare earth elements that are going into some batteries. But I haven’t seen it restrict the availability of batteries. Batteries, right now, happened to be much more obtainable in a shorter time frame than a three- and four-megawatt engine.

Matties: Well Mike, I’ve found this to be really enlightening and I’ve enjoyed this conversation immensely. Thank you so much for your time.

Feinberg: Nice meeting you. I appreciate the answers. And I also appreciate the education.

Matties: Mike, thank you again, sir. We’ve appreciate this immensely.

Mosman: All right. Any time that you guys have a question regarding data centers, mission-critical facilities, or anything like that, you can feel free to give me a call.

• < Previous Page