Data Center Infrastructure Design Webinar l IEEE LAU Student Branch

[00:00:00] Shaheer Shaaban: Thank you, Karla, and thank you for the IEEE Lebanese LEU student branch for inviting me and hosting this event. First, a little bit about myself. I am Shaheer Shaaban. I graduated in Computer and Communications in 2004. And then I worked in the field, mainly in the UAE in Dubai for 14 years almost. And then recently I relocated to Canada. Mainly my expertise, you know, you study something and when you go to the field, you will work in a more specific, if you want, field or environment. So I mostly work with structured cabling. And when I say structured cabling, it means copper data communications and fiber data communications and also data centers as well, because you see a lot of communications in data centers. So in the last five years, I was also highly involved with data centers. Today, our presentation is about data centers, our data centers infrastructure, and I will be talking about components of in data center other than data and IT, because people who hear about data center, they think it's only IT and data, but no, there is a lot beyond data and IT that keeps the IT equipment running in a data center. So before I start, I have my contact details on the screen, if you would like to connect with me on LinkedIn and give me your feedback about this presentation, or if you have any questions, I have my email or on LinkedIn, I would be happy to connect with you all. And I also have a YouTube channel where I put some technical videos, purely technical videos, so if you're interested, you can also check my videos there. So for today, we will be talking mainly about data center types at the beginning, so we'll discuss different types of data centers. We'll talk about the main components in the data center and the systems and solutions in the data centers, and then we'll touch a little bit on the standards, just to see what are the standards related to the data center, and then we'll discuss the different tier levels of data centers. And this is important to understand if anybody is talking about data centers, it's the global classification of data centers. And then we'll talk about some interesting aspects of data centers, just an overview. So enough to cover the one hour that we have for today. Now, if you have any questions, you can write them down in the chat or somewhere, and we'll try to answer them at the end, or we'll see how much we have time. Okay, so coming on to data center types, we have four main types of data center. You see five on the screen, but there are four main types, and we'll talk a little bit. So we have the enterprise data center. We have managed services data centers, and we have co-location data centers, cloud data centers, and the edge data centers. We'll talk about each one in detail and see what does it mean and what is the definition of each of these types of data centers. An enterprise data center, as the name denotes, it's a private facility built for one client or for the owner. So if you're a company and you need to host your data and IT equipment, you could build your own data center and manage it and run it. And in this case, this is an enterprise data center that is privately used by the owner. So in managed services data centers, which is another type, and here the data center is managed by a third party or a service provider, let's say, who rents out the equipment in the data center. So instead of, for small companies, instead of building their own data center, they can go to a managed services data center and rent out equipment, one server, two servers, and then use them while they are hosted and managed by the data center operator. So in a co-location data center, what you rent is not just the equipment, you rent space inside the data center. So somebody will build a co-location data center, and this will be a huge white space, we call it. And then inside this data center, they can divide it using cages or individual racks and rent these spaces to different users. So you can go there, you can put all your hardware, equipment, and use it. And then, of course, you need to depend on the owner of the data center to give you the telecom services and the internet services. The fourth type is the cloud data center. We hear a lot about cloud, and it's the same meaning of the cloud that we hear about. So a cloud data center is an off-premises data center, same as the managed services and the white space. But here, you rent virtual assets. You don't actually rent a hardware or a space. You rent the storage. So you want 10 gigabits per second, 10 gigabits of storage, you can rent. It's similar to when you buy storage on one drive for your personal use. You are using a cloud data center. You can also buy applications with it and software and things to manage it. But for you, you get only the service. You get the storage, and everything else is managed by the data center owner. So Google, Microsoft, Amazon, they do a lot of cloud data center as well as managed services, data centers as well. Now, edge data centers doesn't fall in the same type category of the other four. But I put it here just to explain the concept of an edge data center. So what does it mean? An edge data center is a data center that's placed close to the application, closer to the user. And this is required a lot in IoT applications, Internet of Things, where you have a lot of devices connected to the network. And you don't want these devices to be connected directly to an enterprise data center that is way far from the application. So you bring the data center to the edge, close to this equipment. What is the benefit of a data center? Two main things. One, it reduces the latency. So you have faster communication between the components and the data center. And in case of you have a disconnection from the main data center, your edge data center can continue to operate and manage the terminals connected to it. Now, let's go and look at the main components of a data center. And as we all know, data center is mainly about IT. So when we talk about data center, the main purpose of a data center is to have storage and processing facilities. Someone is talking who's not mute. So please, if you can just go to mute. Thank you. So, as I was saying, in a data center, the main purpose is IT. So the main component is the basic IT infrastructure, which is the racks, the fiber and copper connectivity, and the servers in the data center. And this will occupy the biggest area of a data center. And this is the purpose of a data center. Now, when you have IT equipment, you need to power them. So the other main component is the power. Where you will have the transformers, the power panels, the UPSs, the fueling system for the powering system, and the generators. So this is another main component or system of a data center. Now, because you have power and you have a consumption of power, you have heat generated, and then you need to control the heat. Then you need cooling. So cooling with the chiller plants, pumps, different types of cooling. We will talk about the different types of cooling later. But cooling itself is the third main component of a data center. So these three main components are the basic or the main pillars of a data center. And then to complement the data center, we have also other systems that come along. First is the physical equipment and solution. And here you have your raised floors, your walls, the building itself, the type of containment you have, the grounding and bonding. So all the supporting physical systems for the power, cooling, and IT. And you have the security. And it's very important because everybody is always cautious and worried about the security of their data, about their personal information. So we have two things to protect data. First one is security from a software perspective or from a network perspective. So we have the network security that is not our discussion today. That is part of the data and IT. But what we're talking about is the physical security where you, how can you secure the parameter of your data center and the access to your data center. And usually we have four levels of security. We have the parameter security where you monitor the surrounding of a data center for any suspicious movement. You monitor the facility itself and you control access to it. And then the computer room itself and then the cabinet, each cabinet is controlled by itself. So not everyone going into the facility can enter into the control, into the computer rooms. And in the computer room, also the access is given for specific people. So if you're somebody responsible for the network, you will have only access to the network rack. If you're responsible for the servers, you will have access to the server racks. And all this is controlled in a data center. Of course, safety is also another major thing in a data center. And that's why fire detection and fire suppression systems take a major part in a data center design. So we have different types of safety requirements in data centers, starting from fire alarms that will detect a fire or a smoke. We have very early smoke detection systems that come like tube types will be going all over the data center and sensing for any smoke, similar to oxy reduct. See, it's also an oxygen sensor that also have ducts going all over the data center and it's continuously sampling the air for oxygen. And if it detects a reduced oxygen at one location, this means there's a fire or there's something wrong. It initiates an alarm, of course, the FM-200 and NOVIC-1230 systems are fire suppression system that work on spreading a gas. So in case of fire is detected, these are suppression system that will release a gas that will absorb the oxygen from the air and then suppress the fire. And fire stopping is mainly in the doors and on the openings coming in and out of a data center. So we have fire stopping between connecting areas, between areas of a data center. So if you have a data center with different areas, you will have fire stopping between these areas. So in case we have fire in zone one, the fire stopping will prevent this fire to go to zone two. And we have cables going between these two zones. Then we have to make sure that all these cable openings or the cables themselves are designed with fire stopping in mind. Now, the final main component is the DCIM, the data center infrastructure management. Because you have all these systems in the data center, you need to manage them and control them. And the DCIM provides an overall control system for the data center. Now, data center, like everything today, is also governed by standards, by international and national standards. And if you look at the standards that talk and define the requirements in a data center, you will find we have ISO, the international standard of organization. We have the ITU, the IEEE, Big C, NCTIA, and Uptime Institute. Actually, Uptime Institute is not a standard body like the others, but it's an institute that develops guidelines and services for the data center. So Uptime is one of the main organizations in the data center industry that defines these standards along with ISO and TIA and other standards bodies. So today, in our discussion, I will be mostly referring to Uptime Institute and most of the things that I will refer. Now, note that there is no contradiction between the different standard bodies. They complement each other. So if you look at the TIA, they will find, or ISO, IEC, or TIA, you will find they talk about more details. They talk about more specific requirements in cabling and classifications of a data center. Uptime talks more on a practical approach on when you go to the practical life to build a data center and to manage and operate a data center. Then Uptime gives you a more tangible approach to the standards or the information that you're looking for. Now, so as we said, standards are important. And why they are important? Because our goal in a data center is to reduce our downtime. Now, based on a study published by Uptime Institute Journal, you can see that the different causes of outage and downtime in a data center are UPS failures. UPS, human errors, water and heat cooling, generator failures, IT equipments, and other. And we have 12% from weather related. And you cannot control the weather. However, the other things, you can have control or influence on them. And that's why we can know that 88% of the outages could be avoided. Okay. Or we can take procedures or considerations in our design to reduce the outages coming out from these 88%. And we will see how is this done in the tier level categories of a data center. So, data centers are defined by categories, by tiers, sorry. So, we have four tier levels or classes. So, in Uptime, they call them tier levels, tier 1, 2, 3, and 4. And they're called the classes in TIA or in ISO. Now, so, we have four tier levels. We have no fractional tier. So, if a component is or a part of the system is tier 2 and the other parts are tier 3, then it's a tier 2 system. And always, the classification is based on the lowest system. Now, these classification or the topology categories are defined in two types of redundancies. One is the component redundancy, the capacity component redundancy. So, measuring how much you have of the load. If you have N load, N is the need. So, if you need, let's say, one megawatt, and then your redundancy could be having two megawatt supply. Then this is redundancy in the capacity components. The other redundancy is in the distribution path. How you deliver this component or this capacity to the IT equipment. Because if your power supply fails or the cable supporting or connecting the power supply to your IT equipment fails, you will have the same result. You will lose power and you will shut down. So, redundancy is seen as two parts. One on the capacity components and one on the distribution paths. Let's have a quick look at the tier levels from tier 1 to tier 4. So, tier 4 is a fault-tolerant design. And it gives you 99.9995 availability. And so, in tier 4, of course, when we will look at the design later, you will see that it costs much more if you want to go from tier 3 to tier 4. It's not cheap, it's very expensive, but it gives you almost full-time availability. And this is based on having redundancy on the distribution paths and on the capacity components. Tier 3 data centers are called concurrently maintainable. That means you can maintain the data center or you can work on the equipment or on the capacity components or the paths while the data center is running. You don't need to shut down any part of the data center if you want to do any maintenance work. And this is tier 3. And it also provides a 99.9 availability. See, I have an additional 9 here. It's only 99.982. So, and tier 2, it is a redundant system. It only has redundancy on the capacity component. It also gives 99.7 availability. And tier 1 is a non-redundant solution. So, it's just a normal data center. I will give an example only on one component on the power path for the tier levels to try to understand what the different tier levels mean from a perspective of only one system. And I'm taking here the power or UPS system. So, if you look at this diagram here, this is a diagram of the power system of a data center. You see, you have two power sources. You have utility and generator. You have ATS and automatic transfer switch. And then you have UPSs and going to the data rack. Now, in data center, we use N a lot. Like N equal to UPS modules. And N means the need or the needed power or the needed energy for your data center. So, if we have N equal to UPS modules and I have only two UPSs, then I'd have no redundancy. And I have a single distribution path. Then this is a tier 1. There is no redundancy, not so ever. Now, looking at tier 2, we said we have redundant components. And here, also N is two UPSs. So, we have the same need. We have a single path of distribution. You see, there is only one path here and here. Now, on the path, even if you have, like here, you have three paths. And from here, we have one. So, at any point, you have a single point of failure in the path. And then this is considered as a single path. So, we have redundancy. If you look, it's 2N. We have three UPSs. While the requirement is for two UPSs. So, here, we have a redundancy of N plus 1. Okay, so, in case we have the data center, just to give you more realistic understanding of what I mean here. If N is 100 Watt, okay, let's take it very simple. If N is 100 Watt, the need, we need two UPSs. Then each UPS is 50 Watt to support the system. So, instead of having 100 plus 100, or I can have a 100 Watt UPS, only one unit, or two UPSs, each is 50 Watt. Then I have a 100 Watt. Now, why I go for two UPSs of 50 Watt? Because when I want to make a redundancy system, I don't need to make 2N. I can make N plus 1 in this case. I can have 50 plus 50 plus 50, 100 is the need. So, I have two UPSs serving 100 plus one additional UPS. In case of a failure in any of the UPSs, this one will support the system. Now, if you notice here that in case of any failure in this distribution path, I cannot maintain it without shutting off the data center. And this is an advantage in Tier 3, where I have a concurrently maintainable system. Okay. And here we have one active path. If you see, this alone is a redundant path, same as we had in Tier 2. And I have an alternative path that is supporting my system. Now, in case I still have here the same three UPSs that are N plus 1. However, if I have any failure, I have two distribution paths. If I want to replace a UPS, I can turn the power off. Because in the previous case, if I want to replace a UPS and I turn the power off, then the power goes from the, I cannot maintain the data center operation anymore. So, this is a Tier 3, and it's called also concurrently maintainable data center. Now, Tier 4, it's a fault-tolerant data center. And what's meant by fault-tolerant is that no matter what fails in the system, so if you look at this system, no matter what fails here, I will still have N capacity components supplying my data center after any failure. And this is, so of course, you can totally disconnect this side and run with this side. So, these both distribution paths with their components are fully redundant and can be active anytime. Okay. Plus, if you notice, always we have two power supplies or two sources of power. And usually, in the data center, you always have more than power supply source. So, here, in this case, we consider that the need is three UPS modules. So, you have three UPS equals the need. Plus, we have another three on the other side as redundant. And we have two similar distribution paths supporting the data center. Okay. So, that's about the tiers. Now, when designing a data center, I will talk about some interesting aspects. First, I will start by the approach of a designer when designing a data center. How should we approach the design? And then, we'll talk about some components, designs, and options we have for the design. So, the design approach in a data center starts with understanding the business. So, what the business need is from this data center. And then, based on the business requirements, we start the design from that point. Understanding the business requirements, we can understand what type of application they need in their data center. How many users will be using this data center? What's the bandwidth, the speed requirements, and so on. And then, we go to design or understand what would be the IT equipment requirements. Understanding the IT equipment requirements, we can understand what's the physical requirement in a data center. How many racks they need, what's the space they want to occupy, based on the bandwidth they need, what are the cabling infrastructures required to support this IT equipment. And now, once we have the IT equipment, we can know that, okay, with this amount of power used in IT, we can know that we need to maintain a temperature at that stage. So, we need these environmental requirements from safety, from BMS, from cooling, from DCIM. And then, we go and design an energy system or a power system that can support these requirements. And once we have all this, then we look at the site. Where can we occupy all this material and how can we manage it? Can we put it in an existing building? Do we need to build a new data center? And also, in selecting the site and location of a data center, there is a lot of requirements from the standards, from Uptime Institute, and from the standards. Now, if you're building your own data center, you're the owner of the company or you have a huge house, you want to build your own data center, nobody forces you to follow the standards or to abide by any of the Uptime Institute requirements or ISO, IEC requirements. However, if you're building a data center to serve a client or to sell services, then these have great value. And then, you need to follow the standards step by step and make sure that your data center can be certified later to be to the tier level that you are designing to. So, yeah, just a point on the site location, and it's important. So, the requirements state that you cannot build a data center, let's say, close to an area where there is floods because if there is a flood and the water comes inside, you will have an issue. If you're building a data center, it's always recommended to put it in the second floor, let's say, of the building. So, there is many requirements depending on the location and the site, and also the temperature of the location or the area you're building data center plays an important role in the data center efficiency and operation. So, my recommendations for a data center design are to follow, first, comply with the standards, as I said, because this is something that will classify your data center internationally. Use modularity. And as I showed you in the UPS, we used two modules, and then we added the third one. Modularity will give you flexibility in your design, and it will give you easy maintenance later in the future and easy upgrades. So, always, it's recommended to use modular systems where possible in data center design and to consider an affordable capital and minimize the life cycle of cost of your data center because when you – this is always, you know, when you want to buy a car, let's say, you think how much is the fuel consumption of the car. So, if you're buying an electrical car, you are willing to pay more, but then you save a lot on your operation cost. And in data center, this is more critical because you would think that your capital cost is the biggest cost. But if you design your data center in an non-efficient way, you will find out that your operation cost So, it's always important to find a balance between your CAPEX and OPEX from a design perspective because the way you design your data center reflects on the efficiency of the operation. Even selecting the location where you – so, if you design a data center and place it in the desert where it's 50 degrees outside or you place it in a place in Switzerland, let's say, where the temperatures outside are the zero range, the one in Switzerland will be more efficient because you can use air from outside to cool your data center. So, but many things are considered, okay, so it's – but these things are some of the things that can play an important role in the operation of a data center. And we always target a compact and efficient footprint and layout in a data center because in a data center, each space inside or each small space costs money. So, if you want to make the best use of the space inside your data center, one note on the side of the efficiency of using the space in data center. So, as you know, in data centers, we have IT racks. IT racks are usually 42 units where we fit our equipment or 48. Now, traditionally, these racks used to host equipment that is 3 kilowatt or 6 kilowatt of power. Now, we see racks of 20 kilowatt of power. And why we see this? Because now the footprint or the size of the equipment is coming smaller. Previously, 10 years back, if you want to store the hard disk of a computer or the hard disk, any hard disk, was the same size as the hard disk today. But the capacity of the hard disk 10 years back was maybe 1 over 100 of the capacity of the hard disk today. So, in the same footprint now, instead of storing 10 gigabits, you can store 100 or 300 gigabits. And, of course, if you have more storage, you have more data access, you have more reading from the equipment, then you need more power. So, the power density in the data center in the past 10 years, let's say, or even more, it was always increasing. So, we have seen an increased power density inside data centers. Okay, so, I'll talk about the cooling in the data center and how to control the environment in the data center. And why do we need to control the environment? Now, as per ASHRI, and ASHRI is the American Society of Heating, Refrigerating, and Air Conditioning. They recommend that the temperature inside the data center of the equipment to be between 18 to 27 degrees centigrade. And this is the inlet temperature. So, this is the air going inside your equipment to cool it. And, I don't know, but in my career, I've seen many data centers, you go inside the data center, and it's chilly, huh? It's like 18 or 17 degrees centigrade inside the data center. And, the owner or the operator would be happy that he has a data center that is, okay, freezing, huh? When you go inside, you need to put a jacket if you stay for a long time. Well, this is not efficient because, as per the recommendation of ASHRI, the incoming air to your equipment can be 27 degrees centigrade. And, if you are in Saudi Arabia or in UAE or in the Gulf, where the heat outside is 40 degrees and you're bringing your internal temperature to 18 degrees, well, you can keep it at 25 and comply with the recommendations and the standards. Then, you are using your efficiency in the data center. And, this is one thing, a small thing, that always improves the efficiency of a data center tremendously. If you look at the equipment, the equipment can operate from 10 degrees to 35 degrees. So, in the range that ASHRI is giving, it's fine for them. So, as a designer, you should always design a data center with an optimum temperature control to give the highest efficiency of the equipment. Because, the equipment also will be more efficient if operating at normal room temperature rather than operating at 35 degrees or 10 degrees. And, a more efficient design from the cost of cooling. Now, in cooling, we have different design or different tools or technologies, let's say, of cooling. So, we have a chilled water system. And, in chilled water system, you have a chiller plant outside the data center. It could be shared for other buildings as well. But, mainly, if it's a big data center, it will be only for the data center. And, it will supply chilled water to the data center that cools the air inside or can be directly to the equipment. We'll see in the next slide. We have the direct expansion, the AX system. And, this is similar to the AC units that you see at the home, the split units. They work with the same technology. All your refrigerators at the home, they work all by direct expansion system. We have free cooling. So, if you are in a country where the temperature outside is less than 20 degrees or 15 degrees, then, you can use free cooling to cool your data center. Of course, if using free cooling, you cannot bring the air from outside into the data center. There is special equipment that will filter the air or will transfer the heat from the outside air with the inner air and transfer only the cooling inside, not the air itself. And, we have the evaporative cooling. Also, evaporative cooling is efficient, but it's not very effective. So, if you have a hot weather outside, evaporative cooling is cheap, but it's not very efficient. So, you cannot reduce the temperature a lot inside. But, it works if you have moderate temperatures outside, you have low power data centers, then, it could be an option. Now, to bring this cooling system into the data center and supply it inside the data center, we have different ways of doing so. First one is cold or hot aisle containment. And, when we say hot and cold aisle, so it could be hot or cold aisle containment. And, here, what we do inside the data center, we separate the zones into hot zone and cold zone. So, if you see in this picture, these are the cabinets, the server racks. And, we have a cold area outside. And, we have a closed containment between the racks with an exhaust from the top of the hot aisle that takes the hot air back to the crack unit or the cooling unit. So, this cooling unit could be connected to a chilled water supply or to a DX system. So, it could be both options. And, this is one way. And, now, we also have in-row cooling. In-row cooling are in-row racks. So, these equipment are similar to the crack unit that you see here. Same functionality, but they fit inside the row. And, they release the cold air directly in front of the racks. So, then, it has better performance. So, in such a system that we see here, the maximum cooling capacity that you can deliver per rack is 10 kilowatt. So, you cannot have more than 10 kilowatt racks. However, in the in-row cooling, you can go up to 20 kilowatt of cooling capacity per rack. So, this increases the capacity of your data center. Because, even if you have space here in your racks and you don't have enough cooling, then you cannot use the space. The downflow, upflow, same. So, this is a downflow. You can also have an upflow where the air will be going opposite ways. And, you use the area under the raised floor for the return air, in that case. And, we have direct to the chip cooling. It's not widely used. However, in some specific application, you have refrigerant liquid coming directly to the processor of the active equipment. It cools it directly. This is the most efficient because you are only targeting, you have targeted cooling only on the equipment itself. However, it is more complicated and more expensive to implement. So, it's rarely implemented, but it's more efficient cooling. So, it's used in maybe specific data centers, not in most data centers. Another interesting thing in the design of a data center is how we manage the clean area inside the data center. Because, we don't want dust to come into the IT equipment. These are very sensitive equipment. And, we want as much as possible to keep the dust outside. So, from the concept of pressure, and air always runs from a high pressure area to a low pressure area, when we design our data center, we try to keep the pressure inside the data center hole bigger than the pressure in the supporting areas. So, in case you open a door between area one and area two, the air will flow from inside the data hole to the supporting areas. So, when you come from outside the facility, when the door opens, the air will flow from inside to outside. So, you always have air supply units that push air inside the data center. And, you have air outlets that are running, extracting the air outside. And, it's always required to have air circulation in the data center. So, this serves two purposes. One is having air circulation in the data center, or air change, plus increasing the pressure inside the room. So, when you come from outside, the dust or the air from inside is pushed outside, not vice versa. Next, we have a quick look at the power. And, we have seen this before in a power system. And, this is a simple representation of the data center power system. You have power supply. You could have only one if you are tier one. But, from tier three and above, you are required to have a utility power plus a generator at the end. You will ask, why do we have a UPS and a generator? Because, the UPS is not an alternative power supply. The UPS is only to support you if there is an interruption of the power supply. However, the generator should be there. And, it should have also a diesel tank that supports the operation of the generator for a specific time. So, the UPS, and here many, I've seen this in my career, that a lot of clients would come and ask you, We need a UPS that would support our data center for two hours. You don't need it and you will never use it. The requirement and the recommendation by the standards is for 15 minutes. And, this is more than enough. More than enough. Because, in 15 minutes, your generator will turn on and it will start supplying power and the UPS will be off. So, in a data center, the UPS is only to support you during an interruption of power. The backup power is from a generator system, not from the UPS. So, you have a power supply for, and usually, another thing, that the UPS, if you have it for 30 minutes or one hour, it's useless. Because, your cooling system is not running on your UPS. Your cooling system is running on the utility power. So, if you're using a UPS and you're ignoring the generator, and you have two hours. After 30 minutes of the IT equipment operation, the heat inside the data center will go above 35 degrees. And, by default, the IT equipment will shut down or you will have a fire or whatever. So, UPS, more than 15 minutes in a data center, is not required, is not recommended. However, in some special applications, still, you see that UPS requirements are for one hour or more in some designs, but very rarely. So, the power from the utility will supply the UPS that supports the IT equipment and the other building services and the cooling system. Now, when we talk about the UPS, there is another option for UPS. So, two options. One is the battery-powered UPS that we all know and we all talk about. We have a battery, and we have, if the power is off, then the battery will supply an oscillator, and it will generate a sign signal that will feed your data center. Now, there is another type of UPS, and it's done dynamic rotary UPS. The difference, so, what we are trying to do in UPS, if you think, if you look from a theoretical point of view, we are trying to store the energy. We want to store energy. So, when we have the main source of energy off, then we can maintain our data center for some time. In a battery-powered UPS, we are using chemical energy inside the power. We are storing the power as chemical energy inside the batteries. And in a dynamic rotary UPS, we store the power as mechanical energy. And how is that? We have, we have the, usually a dynamic rotary UPS is always connected with a generator. So, if you look at the previous slide, the generator will be, the UPS, we will move it from here, and we will link it to the generator. So, we have the generator, and we have a shaft connected to the generator with a big, with a big flywheel around the shaft that is always rotating. So, we have a flywheel that is always rotating. It's maintaining its speed by the power of the utility. And in case you have a power shutoff or power disconnection, this rotation of the flywheel turns on the generator that is already connected to the system. So, the power in the, or if you want, the power in the data center is always coming from the generator. Okay? So, you have no interruption. So, there is two types. One is storing the energy as mechanical energy or as kinetic energy in a flywheel that is connected to a power generator and connected to a diesel motor. So, in case of the power is off, the shaft will keep turning the generator, and it will supply power for at least one minute till the diesel power will kick in and then turn the flywheel and the electric generator. So, this is another type of UPS. UPS, usually, a dynamic rotor UPS is only used in huge data centers, megawatt or multi-megawatt data centers because of, it's very efficient, but because of the cost, it doesn't make sense to put it for or to use it for small data centers. So, we only use it in very large data centers. Finally, I want to talk about the PUE or the power usage efficiency of a data center. And this is like a benchmark of data centers. And when you look at a data center, you always try to evaluate it by its PUE, power usage efficiency. What is a PUE? It's the total facility power over your IT equipment power. So, if you have a data center where you have 100 watt equipment inside, IT equipment, then you measure how much your total power supply. If you have 150 watt, so you have 150 over 100, you have 1.5 PUE. So, for every 100 watt used for IT, you have an additional 50 watt that is used to support these systems in cooling, lightning, security system, software applications, anything that's running in your data center. So, this is the PUE. So, if you have a PUE of one, then this is a perfect data center. You are not using any additional power for cooling or lightning or anything. All the power that is coming to your data center is used for. But this is not achievable, of course. So, also, according to a study by Uptime on the data center, existing data center markets, you see that the data center power usage efficiency before 2008 was around 2.5. So, for every 100 watt of IT power, you were using 1.150. So, you need 250 watt to power your data center. So, it went down in 2013 to 1.65. In 2018, it was 1.58. So, it's now in the range of 1.6. So, currently, these days, our average data center operation or power usage efficiency is 1.6. And that means for every 100 watt we use as IT, we have an additional 60 watt used for supporting this IT equipment, which is good. Because of a lot of losses, you have many things in the facility. So, that is it. Okay. That's it for me. Thank you very much. I hope I did not burden you in this presentation and you enjoy it. If you have any questions or... [00:50:13] Speaker 2: So, guys, if you have any questions, please feel free to unmute and ask. [00:50:23] Speaker 3: Hello, how are you? I'm good. I have a question with regards to powering the data centers. We now saw that the entire powering system is pretty complicated. Is it possible that a data center is powered through renewable energies, such as solar energy, wind energy, or are those technologies are still not able to carry like that much, can't provide that much reliable power? So, is that an option? [00:50:54] Shaheer Shaaban: No, it is an option. And you can see some data centers now, they have their roof fully filled with solar panels. However, until now, I don't know, but from my experience, I did not see it as a main power supply for a data center. The reason is, in the data center, you always need 100% reliability, as we mentioned, you need 99.99% availability in the data center. So, it's used as a secondary power supply. I haven't seen it as a main power supply till now. [00:51:36] Speaker 3: Okay, great, thank you. [00:51:37] Shaheer Shaaban: You're welcome. [00:51:38] Speaker 2: I see some questions in the chat box. [00:51:42] Shaheer Shaaban: The data center provides internet. No and yes, internet is a bigger thing. So, the data center needs internet to run. So, what's internet? Let's define internet first. Internet is the connectivity between all the data centers and yourself and me now. So, now we are using the internet. Everybody connected together. This is the internet. So, a data center without internet is useless and means nothing. And the same thing, an internet without a data center cannot do us any good. Because if you search anything on Google now or on anything on the web, all the information is in data centers. So, they complement each other. The data center is part of the internet. It's where all the data is stored. So, everything you see, when you go open a website, when you go online, when you open your Facebook, when you open any application on your mobile, all the data is in a data center. So, the data center is like the big storage for the internet. So, anything you request from the internet comes from a data center. But to access it, you need the internet. Okay? [00:53:06] Speaker 2: All right. So, the second question, is there a physical recycle bin in a data center or is it simply dumped? [00:53:18] Shaheer Shaaban: Physical recycle bin for what? You mean for the data or for the equipment? If you can... [00:53:24] Speaker 2: I think he means data. Mario, if you would like to unmute yourself and ask or just provide me with more details about your question. [00:53:35] Shaheer Shaaban: Okay. I think you mean for the data that we delete or we put aside. No, it's not physical. The data is not moved to a separate place usually. Okay? I'm not an expert in the data part. I'm an IT infrastructure expert. But I know that when there is a data that is deleted, the only thing that's changed is that there is a flag that this is, okay, deleted. It's noted as in the recycle bin. And in one month, it will be permanently deleted. But it will stay in the same location. The only thing that will change is the sign or the label for the software, not for us. So there will be a small flag there. It will tell that this is deleted and it will be permanently deleted after some time. [00:54:24] Speaker 2: Okay. So there's a question that basically a data center conserves everything deleted from people's devices. [00:54:32] Shaheer Shaaban: I get this question. You know, like if you delete your Facebook account, then they will keep the information. Of course, they will keep it. It's not the data center. You know, it's who is operating the data center. And the big companies who operate the data center or the big giant like Amazon, Google, and Facebook, they make money from our data and from your data. They want to know how you think, what's your style in life, what do you like. So they monitor everything. Everything. Even your emails. You think they don't know. I send an email to somebody private and then the second day I open my Facebook or some website and they get an advertisement for the same thing that I sent in private. And this is, there's no privacy. You think it's, there is no, there is privacy because maybe no one knows what you are looking at. But the machines that are running, the artificial intelligence that are in the data centers running and searching your data, they can see everything you do. And usually, they will not delete any data. They will keep it. They will, they have a profile for you and it will always be there for you. So it's not deleted. I don't think, even if you cannot access it or you don't see it. But, but this is not related to the data center. This is related to who is operating the data center and how they deal with this data. [00:55:57] Speaker 2: Perfect. Thank you so much. Are there any other questions, guys? [00:56:04] Shaheer Shaaban: Welcome. [00:56:06] Speaker 2: Okay. So I want to thank you again, Shaheer, for this very detailed and very realistic and specific presentation. Of course, I will share with everyone the recorded lecture, including you. And I will also send the presentation. Thank you for all the attendance. And we'll see you in a future event. [00:56:30] Shaheer Shaaban: Thank you, Karla. Thank you very much. Thank you, everyone. I hope, I hope I delivered something that is valuable for you. And I don't know. I, I, I wish that I can have your feedback, your honest feedback. You can, you can contact me on LinkedIn. I really like the feedback just to understand how people perceive these presentations. Because, you know, I'm, I talk this, like, I understand every word I say. I know it's common knowledge for me, but some words, you know, could be difficult for someone who is not in the field to understand. Or so if you have any comments or any recommendations or any feedback, please, I would be more than happy to hear it from you. And thank you very much. Have a nice evening.