Different Types of Data Centres–and Why They Exist

0:00 Oh, the money in the data center.

0:01 Yeah.

0:01 Money.

0:02 There's a lot of money.

0:03 Open the doors of a data center and Data center?

0:05 Oh, data centers could Data centers Data center, great.

0:08 Yeah, data centers.

0:09 Data centers.

0:10 They've been dominating the headlines.

0:12 New builds, big expansions, and trillions invested.

0:15 But, what rarely gets mentioned is that a data center isn't just one thing.

0:21 They're typically set up in nine different ways,

0:24 and there are a few unique distinctions that give each one an archetype.

0:28 Kind of like the members of your favorite boy band,

0:30 but with less music and even more fan noise,

0:34 which is surprisingly the annoying sound that started it all.

0:38 In the beginning, uh it was really around

0:41 enterprise data centers were the data centers, right?

0:44 And a lot of the times you used to hear

0:45 about were rooms called like computer rooms or server rooms,

0:49 and there wasn't even a data center, it was a let's put all of our, you know,

0:52 servers inside of one room so that all those fan noise

0:55 and all that heat can be shoved into the one room,

0:57 and we don't have to have that on our office floor.

0:59 Well, that transitioned into, well,

1:01 we want to make sure these servers are running efficiently,

1:03 and this room's super hot, we need purpose-built air conditioning, etc.

1:07 Let's make sure we've got power reliability,

1:09 and all of a sudden you've got the like

1:10 the the primitive makings of a data center, and every enterprise,

1:14 every business in the world kind of ran their own

1:16 set of servers inside of their own office building.

1:18 Like John said, enterprise data centers are where it all began,

1:22 the founding member of the band, so to speak.

1:25 These are run by one company for its own use.

1:28 Think banks or big retailers.

1:30 Since they want to keep their data secure

1:31 and meet all of their regulatory compliance rules.

1:34 While enterprise data storage may have started small,

1:38 today these operations tend to be bigger,

1:41 and that's in part because there are more options now,

1:44 like colocation data centers.

1:46 These are places that rent out space

1:48 for companies to put the servers that they own.

1:51 A great choice for medium-size businesses.

1:54 For a lot of companies nowadays, you know, that when you're a small like,

1:56 you know, you're a startup or you're you're a small enterprise or like,

1:59 you know, a home business or like

2:01 that, a lot of folks just use the cloud, right?

2:03 But once you get to the point where it's like,

2:04 "Ah, I I've got probably, you know, a five five-digit kind of monthly cloud

2:10 bill or something like that." That, you know,

2:11 you start thinking buying a server isn't that expensive.

2:14 Let me see Let me look at the economics

2:15 of that and and maybe become a colocation customer.

2:18 Colocation data centers help an organization save money

2:21 while allowing them to keep control over their gear.

2:24 That means, though, that the customer is still responsible for configuring,

2:28 maintaining, and upgrading their own hardware and software.

2:31 Now, if that sounds icky,

2:33 you may prefer the next level, managed services data centers.

2:37 In these, a service provider, like Equinix or OVH, not only houses your rack,

2:42 but also handles the day-to-day running and upkeep of your infrastructure.

2:46 But, of course, it'll cost you.

2:49 Cloud data centers are the mysterious bad boy of the data center boy band,

2:53 remote and occasionally unavailable for reasons no one fully explains.

2:59 They're so misunderstood.

3:01 Who is he?

3:02 These data centers are all about giving you computing power over the internet,

3:06 handled by enormous providers like Amazon Web Services or Microsoft Azure.

3:11 Not only do you not own the computer,

3:13 you might not even be using a whole computer.

3:16 If you think of a cloud data center,

3:18 uh there's again, it's a large centralized facility.

3:21 Uh typically operates 24/7.

3:25 And it's filled with what I would call computing infrastructure.

3:28 Everything from servers to storage systems to sort of the network equipment.

3:33 We literally create subsets of of virtual

3:35 machines in a process called virtualization.

3:38 And what that does is it allows, you know,

3:40 one computer to act like maybe 10 computers as needed.

3:44 At the other time, it allows 10 computers to act like one.

3:46 And that's the notion of elasticity that that, you know,

3:49 we we we we talk about where you know,

3:51 whole unit can can operate as one or a subset of a unit can

3:55 split up uh into a variety of different components and that allows uh you know,

3:59 depending on on the ask or or the nature of of of what's being processed,

4:05 uh it it can it can operate with with with intent.

4:08 Like Chris mentioned, elasticity is a cloud data center's secret weapon.

4:12 It's the system's ability to adaptively allocate its resources as needed.

4:17 Think of companies like Netflix, which mainly uses Amazon's cloud data centers,

4:21 so they can easily grow or shrink streaming capacity

4:24 depending on how many subscribers are watching Love is Blind,

4:28 Secret Island, Getaway, Couples Forever.

4:33 While cloud data centers can be any size,

4:36 they're typically in huge compounds called hyperscale data centers.

4:40 These expansive data facilities are used by giants like Google and Meta,

4:45 and they're also taking off because of the AI boom.

4:49 A standard hyperscale data center typically contains

4:51 a minimum of 5,000 servers, 500 cabinets,

4:55 and spans at least 10,000 square feet, or about a quarter acre of floor space.

5:01 But, they're still getting larger all the time with Microsoft's latest Fairwater

5:05 data center in Wisconsin taking up an enormous 315 acres and counting.

5:11 Bear in mind though that hyperscale isn't

5:13 a completely separate kind of data center, it's a size and design label.

5:18 You'll hear about folks talk about hyperscale data

5:20 centers and you can think about a hyperscale

5:21 data center as a hyperscaler operating a data

5:24 center that's entirely purpose-built generally for themselves, right?

5:28 So, one, they'll either self-build that or they'll

5:30 kind of uh lease that from a provider,

5:32 and they want to have that full dedicated uh region,

5:35 and that could be for their cloud availability zones,

5:38 it could be for AI training nowadays, it could be for, you know,

5:41 their their SaaS applications that they're delivering to customers.

5:44 Uh Uh it could be for ad analytics for instance.

5:47 I mean, Meta's obviously got huge amounts of data center capacity,

5:50 and that's just for them, right?

5:52 They're not operating cloud services for anyone else.

5:54 On the opposite side of the size spectrum,

5:57 edge data centers usually have a smaller footprint

6:00 that fits right in the thick of a business's work.

6:03 But today, they're more often thought of as just data centers

6:06 that are located closer to where people actually use their services,

6:09 like on the edge of a city.

6:11 This lets things happen faster with less latency.

6:15 These kinds of data centers are thriving right now, too.

6:18 Not exactly Harry Styles level fame,

6:21 like hyperscale cloud centers, but definitely in the Zayn camp.

6:25 Telecom companies use them for things like streaming or gaming,

6:28 but they're also powering robotics, physical AI,

6:31 and systems that need real-time processing.

6:34 What we think about as as edge is

6:35 really like kind of like almost a converged edge,

6:38 which is close enough that all of the applications,

6:41 like even things that feel pretty real-time,

6:43 like video and and streaming analytics,

6:45 they can happen within kind of a 10-ms radius.

6:48 That's pretty close to as as as close to the end user as you need to be

6:53 without actually being on the device or like

6:55 on the factory floor or in the vehicle.

6:58 You can probably sense by now that these different

7:00 types can be used in combination with each other,

7:03 like a tight three-part harmony.

7:05 For example, OpenAI generally uses a giant hyperscale campus

7:09 full of cloud data centers to power its AI,

7:12 but they teamed with a company named Armada to deliver edge AI

7:16 to remote industrial sites in form factors

7:18 as small as a single shipping container.

7:21 These smaller units are also known as modular data center.

7:25 Even within a single data center,

7:27 there can still be various types of server racks that allow

7:29 a single building or campus to act in many different ways,

7:33 which brings us to the more niche, perhaps even buzzwordy types of data centers,

7:37 which we'll cover right after this message from our sponsor.

7:40 War Thunder, it's the comprehensive vehicle combat game available on PC,

7:44 mobile, PlayStation, and Xbox.

7:45 There's thousands of combat vehicles for you

7:47 to choose from and pilot on the battlefield.

7:50 War Thunder is known for its authenticity and immersion.

7:53 All of the vehicles you can take control of are

7:56 real-life machines all throughout history from countries all over the world.

8:00 So, join over 95 million other players for free

8:03 today by using our link in the description

8:05 and we'll even hook you up with a massive bonus pack if you sign up today.

8:08 What?

8:09 Yeah.

8:10 Hey, here's a couple new kids on the block getting

8:12 a lot of attention since AI started shaking things up.

8:15 High performance, high density, and sovereign data centers.

8:19 High performance/high density data centers draw

8:22 a lot of energy since they're typically installed

8:24 in giant hyperscale sites and pack as much

8:27 compute as possible in each server stack.

8:30 A recent report from the International Energy Agency states that in the past,

8:34 a typical data center would consume between 5 and 10 MW of energy

8:38 while the new high-density hyperscale projects can push upwards of 100 MW.

8:43 And to put that into perspective,

8:45 a single MW hour is enough energy to supply roughly 300 homes for a full day.

8:51 High-density centers are inherently also high power,

8:54 but not all high-power data centers are high density.

8:58 If you're confused, it's okay, stick with me here.

9:00 Like the boy band references in this video, they're fading fast.

9:03 But one example of high-power,

9:05 low-density would be something like vintage geographical weather

9:09 gear which can draw a ton of power,

9:11 but the hardware needs to be spread out since they generate so much heat.

9:15 Don't expect to hear lots about these unless they're

9:18 being retrofitted to pack in modern servers with more density.

9:21 Finally, there's sovereign data centers.

9:24 These are having a bit of a moment right now and they

9:27 can take the form of pretty much every type we've mentioned already.

9:30 They're popping up as governments and businesses

9:32 realize that to control their data,

9:35 it needs to be stored in a place where their local laws provide jurisdiction.

9:39 That said, this can get murky.

9:41 For instance, US companies operating in Canadian data

9:44 centers are not safe from US government interference.

9:47 For full sovereignty,

9:49 you need the data center and the company operating under the same flag.

9:53 So, those are the types, such as they are,

9:55 but that's not the only way to categorize data centers.

9:58 Another is the Uptime Institute's tier system.

10:02 It ranks data centers from tier one at the bottom to tier four at the top,

10:06 based on things like uptime,

10:08 which means how long the system has been running without interruptions,

10:11 redundancy, and how well they handle failures.

10:14 And basically, it looks at four tiers.

10:15 It looks at, you know, things like your availability,

10:19 making sure that you've got available power that's consistent to process

10:22 up to the maximum that you're you're allowed to process,

10:25 cuz typically there's peaks and valleys in in consumption as well.

10:28 It looks at redundancy, which means if one of these things fail,

10:31 if your power fails, then what's your backup?

10:33 If your cooling system fails, what's your backup?

10:35 If, you know, your part of your IT infrastructure

10:38 or your or your chips and servers go down, what's your backup?

10:42 It also looks at fault fault tolerance.

10:44 So, things like, you know, how often does do these things have errors, etc.

10:49 You know, what's what's in the manufacturer's

10:50 specs and and how are they performing?

10:52 In some ways, I would call it a bit of an audit.

10:55 What Chris told us is basically the checklist data centers live and die by.

10:59 Power, backups, backups for the backups, and then backups for those, too.

11:03 If you want to get a little more up close and personal with a data center,

11:06 check out this video of Linus touring one.

11:09 And drop a comment if you can guess what type it is.