Datacentres

This time we're going to talk about data centers. Data centers is one of those topics that , all of us have heard off. And even if you didn't, you could probably guess what data centers are anyway from its name. There's so many things I know about data centers. but as soon as I think about talking about something, I feel like all of those different knowledge of data centers comes together in an unorganized way.

In this conversation, I'm looking to gather my thoughts, organize all the different knowledges I have and also learn something new, maybe so thank you for tuning in and let's. talk about data centers.

But I think before we talk about data center, we need to actually talk about how People used to store data, transfer data, et cetera, before data centers. Companies used to manage their own IT infrastructure. And when I say IT infrastructure, that means all the computers that were provided in the offices. So suppose an office is a physical location at one physical location, you have say a hundred computers.

You have to manage, update software make sure they're all secure, they're all connected and the backups are also in place. So that's how it all started from there. if you scale it up, so now you have say multiple locations and you still manage your computers, but each location has its own IT team.

But what if you wanted to have a data center, say, in one city, you have two different office locations or five different office locations, and you want to manage all the IT infrastructure from a central location, and a data center can give you a clue about what it actually is. It is basically storing the data in a central location. That's why it's called a data center.

So what is the deal here?

But if you have a data center where all the backups are stored, then you know that, , in case a computer fails or in case a person leaves or whatever, I think the infrastructure is far easier to manage.

Once we understand what are the different types of data center, I think that would make a lot of sense because I think a data center is not applicable for say a startup. It could be, but in a very different manner. Say you have computers and your IT infrastructure, your server in one central location. And when I say a server, the server is basically hosting a webpage, which can be accessed over the internet.

So you're hosting your server where all the files are stored, your database is there, and people in the office are able to access the information on the database and work on the server and create your website. Plus you have your local machines and these desktop machines are managed by the IT team. So that's one part. Now, suppose you are a very big company, you have thousands of people and you want to manage the data of not just the employees, but also the product.

So suppose you're making a CRM application and you want to manage that. those CRM application has a lot of data coming from the customers and you want to store that data. The data center is not the data of the people in the office, but the data is of the people who are actually using your service. I think the key thing is It's in the word data center. It's a center of data, but there are two types of data in any organization.

One is the data of the employees who are working at the offices and the other data is the data of the customers who are accessing the services of the I. T. Company. So suppose I want to manage my own I. T. Infrastructure. So for that, I don't need to manage a lot of infrastructure in terms of data center.

But if I want to manage my customers data, then I think that has to be managed in a far simpler way because I need to first make sure that the customer is able to access my infrastructure securely, safely and very quickly. So they get access to that information. Plus they can get access to that information 24 7. So it means it's always on plus it has to be secure so that no one is able to hack into that data and leak it onto the internet. So that's why it makes sense to have a a data center.

This data center could be managed by the company. So suppose I build a CRM application, so I manage the data center. I take care of the physical space in which the servers are located. So there are two aspects to it, right? You have to manage the physical space. So people can't break in and the computers will get very hot so you have to provide some cooling, you have to maintain the humidity, you have to provide access control. So no unauthorized person can access the data.

And you have to make sure that it's compliant with all the regulatory bodies. And then you have the physical infrastructure where in you have the computers, the networking equipment and the servers. So that's part of a data center. Now, in a different data center, the physical space itself is managed by a separate company, but the equipment, computer equipment is managed by the company itself. So suppose you are a company and I manage the physical space.

You say, I want to rent your space a hundred square meters and I want to put my network servers there. I'll provide the security. I'll provide the cooling, but you bring your computer equipment. So that's co location data center. The third one is where I give this physical space. And I also provide the physical infrastructure of the computing equipment. So I'll provide the computers, I'll provide the space and then you can use that as a data center.

And the final is cloud service provider like Amazon web services. Google cloud or Microsoft Azure. And these are infrastructure that is all virtualized, which can be accessed from anywhere in the world. So these are the four major data centers.

While you were describing all of these, I was also thinking that what are the motivation behind having each of these types and, going back to the first question I had is, why are we, increasing risk by having everything centralized? But then I thought it's all about isolating the problem and finding the solution for that particular problem, when someone says, don't put all of your eggs in one basket, they are trying to say that, okay, If you lose the basket, you lose all of your eggs.

So how does it translate into data centers? If you put all of your data in one place, and then if the physical location is subject to an earthquake or a fire and you lose all of your data, so that's one problem. And there is a solution for that problem, which is having backup, online backup, offline backup, very quickly retrievable backup. And the same way there's all of these other problems that centralizing data comes with, there are isolated solutions for all of those as well.

Ultimately, it's a addition and subtraction of what you gain from it. So you save a lot of cost by centralizing it. And even having two or three different backups, you would still save money by having a centralized location for all of your data because it would just reduce maintenance and management of it as well. So you would have to hire less people to manage it all together. And that's a lot of money as well.

It's not just the infrastructure management, but also a huge team of solutions architects and data architects who would, make everything work together. After all of those calculation, it's more worth it to have a centralized data center, and that's why we have them.

I graduated my university in 2006 and that's when Facebook came. It's 2024. So it's 18 years since I graduated. It has stored every single like every single photo, that I've clicked on every single comment. And there are 1 billion people accessing it. So imagine the amount of storage in terms of images, files, et cetera. Now, if you have to store that physically in one single location, of course, that can create a failure because if that place crashes, then all the data is lost.

But when you create a data center, you have to look at what, The end goal is if I'm a very small company, I don't need a big data. I don't even need a data center. I can manage the infrastructure in my own company. I just need a small team to handle the maintenance of the servers to update the security patches to control the networking equipment and to have multiple hard disk in a RAID format. So RAID is basically to provide redundancies for hard disks.

So basically if you have multiple hard disk, you can put them in different RAID configurations so that it can create automatic backups without you configuring it regularly. So that's one part. But suppose now you have A million customers. Now you can't have that server in your office because it will require a lot of space. plus you have to manage the cooling, then it's better you move to a separate physical location from your own office.

So that way you can minimize the cost because now you have a location just for optimized for data center. you don't need tables. You don't need chairs. You just need a very big empty space. You need a good air conditioning. You need humidity control and you need people to guard and secure that location. So it's a very simple box in which you stack a lot of servers. If I want to now scale my infrastructure because I have that big empty space, I can scale it up very quickly.

But if I have the servers in my own office, I need a different room. And then I have to figure out how to connect the servers into different rooms. But if it is just one physical location, it's just like attaching one cabinet, putting the hard drives, putting the network cable and then connecting it. And then it's all online, right?

If it's just one branch and you're starting up, it makes more sense to prepare the burger patties and all of the individual items in your kitchen and then prepare and serve the gourmet burger and that's your business, but as your business grows and then you compare, start comparing it with fast food chains like McDonald's and the similar, where, you have grown to a degree that it no longer makes sense to create the burger patties at your location.

You probably have a separate central location where you create the burger patties for five or ten or a hundred restaurants in the area. it also could come to a point where you outsource or delegate to another company who are better suited to cut and slice the potatoes into fries. Even though you are a fast food shop and, you don't think that just because of that McDonald's should, prepare all of their food by themselves.

No, as you scale up the sort of the load of work becomes so high that it makes more sense to delegate or give that particular part of the task to a different company who would just do that and supply that. Even in IT company, you could specialize in say, for example, social media being a social media platform like Facebook. So your expertise is not to manage data in cloud, . So you might want to hire, I don't know what the actual configuration is Facebook's backend, but you might want to.

Microsoft Azure or Amazon Web Services, who would handle all of your cloud infrastructure. And that way you can focus on what you do best. In case of Facebook, I don't know actually what they do best. But whichever IT company that we can talk about whatever their expertise is, they can focus on that and they can outsource or delegate this part of managing data which is a very important part. But, as it scales up, it becomes a very big challenge to handle by themselves.

And that's where cloud data centers comes in.

So cloud service providers, provide access to a networking and storage solution over the internet. instead of buying multiple servers you basically rent it. And you say, I will buy a portion of your infrastructure with this processing power, this CPU, this storage etc. And I will rent it out for a certain period. And I will pay a subscription cost for it. They can provide the infrastructure plus the configuration. So they'll manage the firewall, they'll manage the DNS servers.

You don't have to do the configuration. They will also install all the softwares. So you have the physical hardware, the hard disk RAM and the CPU. Then you have the configuration. You have to set up a firewall. You have to set up a DNS. You have to set up the server software and a database. So they do that. And then you have the software to manage all this, right? So there are different levels of what a cloud service provider can do.

And based on your requirement, you can then say, okay, I want to pay for this and I will manage it in my own way. I'll configure it in my own way, but I'll rent it from them. And because I'm renting it from them, I don't have to worry about the physical security. I don't worry about cooling. I don't have to worry about paying the electricity bill.

I don't have to worry about securing the physical location with say security staff, and then I don't have to worry about updating the infrastructure because I'm now getting the latest CPUs and the GPUs, which I need. So that's a benefit of a cloud service provider. Of course, as I mentioned, you have managed service providers, co location and companies having their own data center. So sometimes when you are such a big company, it makes sense to have your own data center.

Amazon or Microsoft may be having their own data centers.

And this single hardware can serve 10 different people. So that's the virtual machine. Now, when Amazon became such a big company, it had to create a big data center to manage all the shipping orders and the logistics and the data that is actually created to create the whole website, right? So it has to manage all the images, it has to manage all the pricing, it has to manage all the authentication, login, etc. And plus store all the orders.

And when it started growing big it created big data center. And then it thought, okay, I am managing all this data center and it's not getting fully utilized. Maybe I can rent a bit of it to other people. That's how cloud service started.

They ended up creating something which was really, necessary for a lot of people. So they ended up selling that. it's very interesting how you come up with new products and there are so many more opportunities coming every month, every year. People might think that. I. T. Has oversaturated with all of these new businesses, . But there is new inventions coming on every few days.

And right now we're seeing a boom, with A. I. And new products will be made. Cloud services and all of these things. All of these were possible as the Internet became, more commonplace. That's why you could think of, okay, I have broadband, I have MBPS, not KBPS, so I could stream data, I could, collaborate online with multiple people. All of this became possible as the Internet speed grew.

So now with AI and large language models and content generating models, which generates video, pictures, audio, all of these things. Now we're going to see, very soon , a new type of products that are emerging that people didn't even think about. It's an exciting place to watch the space and also think about yourself.

Do something, maybe the thing that you set out to do that, that might not be the thing that makes you successful. But, in order to make that possible, you'll probably build something else around it. Which will then become the thing that makes you known and successful. I think one of the things that I always encourage everyone is to basically do something. implement something, take action on anything and you never know how it will turn out.

We have talked about data center, but let's talk about the challenges of the data center and maybe what the future lies ahead for data centers. So one of the biggest challenge for opening any data center is not the hardware, not buying the software or the network switches, routers, etc. It's basically the energy. They consume a lot of energy. So you need to get subsidized electricity.

Or subsidized energy in a form that is cheap for you to operate it plus because they consume a lot of energy to operate, they also generate a lot of heat. So you need a lot of cooling. So actually, data centers main cost comes from these particular things. So you have energy and you have the cooling and then you have to maintain the humidity because you don't want any computer infrastructure to get spoiled because of humidity.

So you have to manage the energy, manage the humidity, manage the cooling. So once you've managed all that, you need to manage the physical location with security personnel, provide access control, et cetera. Plus you have to update the hardware because hardwares do fail because they are constantly running at a very extreme temperature in a very sensitive environment that in case one fails, you have to immediately replace them. Now, of course, all these servers have their own backups.

In the same location or in different location in a different data centers. So if you think of AWS or Google, they don't have one physical location where all the data is stored. They have multiple data centers across the world. So they have one data center in say UK, one in Europe, one in US, one in Australia. So suppose you are based in Europe. They're going to serve the data from Europe servers, not from America, because the latency will be very high.

That means that the time it takes for the data to travel from US to Europe is quite high. So they would rather serve it from a location closer to you so that They can instantly serve you the data, plus it creates a backup. So suppose there is a tsunami in Australia, then the data stored in Australia is also backed up in, say, U. S. and Europe. So not all the data is lost immediately because they have redundancies in place in terms of different physical location. So that's one part.

So you have to manage the heating and to generate the energy The couple of option is you can go for generating your own energy. You can have solar cells but the solar panels will not generate enough electricity. So you have to maybe burn fossil fuels or you have to go for a nuclear power plant option, et cetera, et cetera. So it depends on how sustainable you want to go.

And a lot of companies now they buy carbon credits and they say that, okay, our data center is completely carbon neutral because whatever, emission they produce, they are neutralizing it through some other means.

We can replace that with robots so they can retrieve the data for you instead of humans. So that's one way where the data centers are heading.

The other way is Using AI and machine learning. So you can use the algorithms to predict when the next hard disk would fail or when the next memory would fail. And you can based on that prediction, you can order the spare parts and be prepared to replace it as soon as it fails. So you have less downtime and you have instant connectivity. Then quantum computing is now being built. So how do you create a data center that has quantum computing provided to all your customers.

So you have to think at that level. At a high level These are the challenges. And these are, this is the future in which the data center is heading.

But I think one of the main things that you might have noticed, especially with apple intelligence getting launched recently the focus is AI running locally on your own hardware, right? So the data has to stay on your hardware. The algorithms have to run on your hardware. So in an age where data itself Is being managed on your edge device, which is called very close to your location or basically in your hand edge device could be your mobile phone. It could be a thermostat.

It could be a doorbell. So these are edge devices, right? They're all connected to the internet, but they are actually closer to you than a data center. And the data is normally processed on board that computer and then stored at a different location, but it's processed there. So what happens to those edge devices? And because there are so many edge devices now or you can say internet of things, IOT devices, then where does that data go?

You may not want to store all the information or all the logs in a data center. You might want to store it in your own house. So you might have your own backups or systems in place to do that, right? That information you don't want data center.

That's whether, moving from Mbps to Gbps internet speed is a better option than localized processing. Because if you want to do localized, You have to think about the cost of hardware that needs to be present and that's the whole point of I. T. and, making everything remote that, make everything more efficient by having one thing in a central location, which does, which serves like 20 people or 20 devices. But if you need to do localized processing, that's a totally different story.

Obviously it could go both way. We never know what the future holds. But, as you're saying that one of the options for data center is to move that way. I'm thinking that maybe it's more the kind of the overlords of the I. T. It might actually want to focus more on how to improve that communication so they can still do the processing remotely. The data transfer is so high speed that it feels instantaneous.

They are not stored on the, say Google drive or Google photos. I'm talking about an Android phone. So, that processing is local. Plus you have certain emails. So My emails are stored in a data center, but the AI processing is happening locally. So now there are two aspects to it. The data itself that is being retrieved from a data center and the processing. I don't want the processing to happen in a remote location. I want the processing to happen in this location. Why?

Because the moment the AI algorithm works on my personal data, it trains itself and I don't want it to train on my personal data. I want it to be trained on public information, and this is private information, right?

understanding of all different kinds of email, all different kinds of people sent.

And now someone who knows that this is the training data it has been trained on, you can actually figure out a way to retrieve that information from the algorithm, right? So you don't want that. So that's why when you talk to chat GPT, you don't say that, okay, this is my company. I work at say Amplify and I do this and this is my current problem. This is my boss name. I want to create an email.

No, you don't want that information because that sensitive information, how do you think is getting anonymized? I don't think that if there are billions of people putting inputting all that information, do you think there are billions of people trying to anonymize the data? I don't think so.

So they're optimized just for those types of tasks. And these silicons are now being built and they're getting installed on the smartphones or the laptops. So even in Intel series, you have the Intel Core series and the Intel core ultra series. The ultra series has AI cores in it and then the latest M1 chips from Apple, they have those chips as well. So the Apple intelligence won't work on say anything below Apple iPhone 15 or any Mac book that doesn't have an M1 chip or a lower chip, right?

So it will be limited by the hardware infrastructure. But over a period of time, that infrastructure cost will come down and it will become more accessible. And then more and more computation will be done on your devices. Even Google for their pixel devices, they manufacture their own tensor processing units. So I think there is definitely a possibility of Scaling it

And I hope our audience would continue listening to different topics that we talk about every few weeks. One of the things I have heard at one point is that Microsoft is thinking of putting their data centers on the seabed because it's so cold, it would just be naturally easier to

So how do you manage data centers, say, in a world where you have moon bases or your bases in Mars, so that will provide a different set of challenges. Now you have to provide communication between Earth, Moon, Mars, et cetera,

So that happens in Antarctic because they get internet only for a few hours a day when the satellite passes over them. And that's how they get the internet connection. And now it's very slow. So in that slow internet speed, what infrastructure can you operate and how can you manage that effect efficiently? I think that is another type of challenge, right? It's because we're so used to high speed internet.

We think all problems can be solved with providing more bandwidth, etc. But, there are locations where you just cannot provide high speed. How do you manage infrastructure and how do you set expectations for that kind of environment? And I think that is again a beautiful challenge to solve.

The same way when we talked about cloud in one of our previous episodes, we talked about cloud computing, cloud infrastructure, even though the name makes it sound like it's somewhere over there in the sky that we don't have to know about, or even it's, it may be virtual.

But actually, It's quite closely related in this way that , even though we say it's in the cloud, it's probably in a data center somewhere, which is a physical location, heavily guarded, heavily, protected and environmentally controlled in a granular way to make sure that everything's functioning. this physical location could be anywhere as we've discussed today under the in arctics or in, in countries where electricity is probably cheaper.

Many things to consider when you are thinking of data center, but the necessity of it is definitely something that is unavoidable for small companies and for large I. T. Organizations I. T. Or not even for banks. They have their own data centers a lot of the times for the very reason of security and privacy off of their customers. Whether you're an I. T. Company or not, as long as you're, a mid sized company in any industry, you would probably need data center.

to be able to operate and provide services to your customer, and that's why data center is a very crucial bit of technology that is everywhere and is needed by everyone as anyone is using some sort of technology on a day to day life.

Anyway, so it's a very sort of important topic to know about, learn about, and also to train yourself about, because as people are talking about, AI taking over people's jobs and which skills are becoming out of date, etc. Anything to do with data centers as it may evolve.

The technology, but knowing about data center being trained and certified on various aspects of data center will always be in place because, AI or any other technology can't survive without data and it being transmitted very quickly. So that's why it's very important.

It requires a very large amount of electricity and that electricity can be provided through maybe hydroelectric, nuclear, fossil fuels, et cetera. So it creates its own challenges now, especially with the infrastructure because you have to train these large language models with tons and tons of data.

Tell us what your idea is. And we could say, okay, this seems plausible. You should, pursue it more. So it would be an exciting thing to do. we look forward to getting your feedback and communication on how we would solve a lot of these challenges that we talked about today Thank you very much. Very much enjoyed talking about data centers. Hope our audience did the same. We look forward to getting your feedback and to hear from you. with that thought I'll end here.

Thank you everyone for listening.