Cooling Strategies for Data Center Design and Energy Efficiency with CFD (ASHRAE 90.4)

[00:00:00] Speaker 1: Hello and welcome to this SimScale webinar on cooling strategies for data centers welcome everybody this nice European evening and good morning to all our listeners in the US just to make sure that we are actually right now that you can properly hear me make sure that please raise your hands that so we confirm that you can see the screen and that you can hear me right now so really appreciate if you could use the go to meeting functionality before we begin to check if you actually hear me speaking and if you can see the screen I still do not see I see a lot of hands appearing this is great thank you very much great for those of you who cannot find the there's a possibility of listening using a phone let me just type this in the chat all right the rice hand option I will be used throughout this presentation let's just before moving on with all the talk let's agree on some basic errors I will try to answer questions as they arrive throughout the presentation although I would highly appreciate if you could hold with them up until the end so that while we get like a nice workflow and a nice thought pattern and that is going on throughout the presentation all right with no further ado let's jump in once again welcome to the sim scales webinar on cooling strategies for data centers my name is Pablo Sosnowski I am the customer success engineer at simscale I've been working for quite long over here my specialization is environmental industrial fluid mechanics and I've been dealing with quite a lot of applications of the customers of simscale it's my pleasure to be your host today and I hope that you will enjoy this session if you have any follow-up questions who would like to get in touch with me directly you can find me at this email address [00:02:54] Speaker ?: without further ado let's begin what is why is it are we interested in data center cooling strategies well most importantly it is because cooling takes up [00:02:54] Speaker 1: around 50 percent of the energy consumption of the energy consumption of a data center overall there are many other aspects but this is the place where you can actually have significant cost reductions of your process in fact when we are looking at the graph that tells us about the general usage of energy for big data centers the energy consumption of a data center overall there are many other aspects but this is the place where you can actually have significant cost reductions of your process in fact when we are looking at the graph that tells us about the general usage of energy for big data centers we notice that well basic interpolation of the trend that we were expecting to have around 2010 was that well it will be exponentially growing and we will end up having quite a lot of trouble fortunately there were some significant improvements when it comes to the infrastructure and the service savings meaning that we improved the ways that we're cooling our data centers we improved the performance of the devices themselves so today the trend is rather flat well it's still rising but it's not as dramatic as it was but most importantly we see two areas where we can actually significantly improve the data set the data center performance that is the server savings so the power is actually being consumed by the devices themselves and the infrastructure savings which is the device that cooled the the tools now there are many aspects when it comes to choosing the location of a data center when it comes to choosing the proper tools proper hardware and so on and so forth um i will not go too much into detail because we are all here to to learn a little more on how the simulations want to go and i'm sure that most of you would be able to tell me much more about data center that we are there are many aspects to that and our comparison or our example that we're doing today we'll be focusing on the approach that is most suitable to facilitate high computing operations but what we are actually interested is are the adapting of cooling strategies within a data center that is actually being built we are there are many aspects to that and our comparison or our example that we're doing today we'll be focusing on the approach of the hot or cold aisle and the containment of such an aisle the benefits that we get by applying such a strategy what does it mean to have a hot cold aisle containment and hot cold aisle structure for the data center well basically we will assume that there will be we will have rows of servers standing in a big room and we'll have consecutive row between the between the rows of structure of these structures we will have consecutive elsewhere from the bottom we will supply cold air and the next aisle that will be connecting two backs of the servers will be extracting hot air and moving it to the recirculation units now this has its advantages well we know that we will not be pushing hot air from one server to the other at the same time well then a naive approach would be well we just keep the building as it is and put the suction but there is a significant chance that recirculation or air circulation within the room will still allow for the hot air to circulate within the within the zone and decrease the efficiency of cooling that we could have if we actually contained the hot aisle or contain the cold aisle and actually we have two possible configurations we could completely close the cold aisle and have the cold air arrive through it not allowing it to mix with the hot part that is around that is being pushed out or there's the second approach where we are containing the hot aisle simply creating a big channel of hot air that is being directly pushed into the cabinet that drives the flow out and processes the air so this is another look at the the approach of contained hot aisle we are not allowing the the hot air to recirculate we are building some we are building some kind of a barrier and it actually doesn't have to be a very let's say thermal resistant one although this might also help if we are really concerned about potential heat losses within such a system now how is it connected to um the the standards themselves what kind of standards are we applying to these um these things one of the standards we're very prominent one is the ash ray technical uh standard 9.1 which would identify the amount of humidity and the temperature that is supposed to be provided to any electronic equipment or it equipment as an input for the air that is ventilating such a such device the standard itself is rather big but we are focusing right now on the ways that cfd or computational fluid dynamics and in general computed aid in engineering can help us cope with the standard in this case we are looking at the performance of the temperature range at the inlet of an iet equipment and we will be considering if our temperature of the air we are supplying to the device stays within a certain range different simulation types allow us to check and confirm that we are within the standard for other purposes or other other properties of the flow in this case we're mostly focusing on the temperature part now why is this important well i would say hey i could push a lot of air a lot of cold air and be sure that well i get a performance that i need and that the temperature is always below the desired 27 degrees and it's not too cold but on the other hand if we are able to properly optimize the system to properly optimize the system itself and the design of the room the design of the recirculation patterns that are within the domain we could significantly reduce the costs the operational costs for a data center some other standards or some other ASHRAE standards that facilitate this kind of analysis are the ASHRAE 90.4 which focuses on the energy consumption and one would say hey the the standards are mostly that you can calculate these standards based on the general understanding of the of the design of a data center data center i have 50 servers i have or 50 stacks of servers i have a room that is that big each one of them provides that much power i have a certain amount of pumps and the data center has a certain energy input but then if we want to optimize it and make sure that we are actually having a proper efficient system that does not consume that much energy and still is within the standard we can use a computational fluid dynamics this also goes to the standard 90.1 where by minimizing the energy or minimizing the temperature that is created or the amount of heat that is being created at the device or by making sure that we have efficient ventilation we are reducing the amount of energy that the it equipment requires or it's that at which it is operating and since we are having the the equipment operating at more favorable temperatures we are getting better performance of this equipment increasing that our the overall performance of the data center and for example the power usage efficiency goes up we get more for less energy that we're putting into the equipment itself well right now let's talk about the design scenarios that i would like to present to you right now we'll have two different design variations one of them will be the naive approach or the old approach where we have our uh racks of servers positioned in these in the six in six rows well there's one two three four five and six counting from the left and there's one small rack in a separated room that you can see and up a little bit up to the front we have three computer room air circulation units or air conditioning units positioned in the back that will be taking the air out of the system and will be uh later introducing it down into the into the base the bottom side of the of the room the air will be entering the room from the bottom uh the the raised floor through uh small openings uh racks that that allow it to to go into the cold aisles a second optimized approach will we will consider having isolated aisles the isolated hot aisles that will be taking away the heat through the top boundaries we will not actually simulate the plenum itself we'll consider the air to be completely immediately removed into the air the air conditioning units and in this case we'll also have different types of optimized flow notice that on the left most hot aisle but when we look at the left most hot aisle it's completely isolated from uh from the flow it's basically we have all the server racks the hot aisles then the hot aisle the second hot aisle looks very similar to the first one although if you have a careful look at the server the server racks that are closest to us you will notice that they are empty let me try to get my spotlight tool i hope that you can see the spotlight we have uh we have empty racks the very front which is semi opening making a semi opening to the hot aisle and will allow some kind of air recirculation and the third type is a rack which is actually not isolated there are suction outlets or open outlets that will be taking the air away from the system although on the side walls they there is a complete opening of this device all right so why should we use the simulations at all in general whenever we are building such a facility we'll start with a certain design we'll build a physical prototype then perform the thermal testing and in case it is not compliant with what we are expecting we will make a design change whenever we're building data centers we have tremendous amount of experience and this allows us to basically follow the guidelines of what we know in order to avoid big problems most of the time if you are trying to comply with actual standards you will end up having a working and reliably efficient data center the problem arises when you actually build the thing and you find out that your servers are overheating it is rather hard to imagine that we'll start rebuilding the whole facility and this is why it is important to have an understanding of the performance of such a system before it is built before we even make the first prototype because in many cases we'll not be able to build the first prototype at all let's take an educated guess we have to base on our experience and although it's very good at most of the times sometimes it might lead to a disaster but the good thing is with the use of the simulations we can reduce the risk we can reduce the number of physical testing that we actually need to perform in order to get a reliable understanding thus we're saving a lot of money simscale will allow you to do all of that in your web browser we're the first world's cloud-based cae platform with full functionality that allows you to run very complex analysis we're accessible you do not need any hardware no actually you do not need to build a data center to run your simulations although for most complex analysis tools you will actually need a very powerful machine to run these problems simscale is cost efficient with a flexible pricing model you get very good performance and very good value for what you pay and it's for everyone there are hundreds of projects already present in our public databases which can be used as templates for your analysis also the projects that we are showing here will later be available for you to take a look investigate change a bit and see how can you apply them to your applications and use them as a template to start and kick off your own analysis of a data center the whole workflow of simscale happens in the in the web browser you upload your can from the very good tool that you're using we are supporting quite a lot of formats right now although we are sticking to what we're good at that is the simulations which is set up in the browser and later you're able either could post-process them online using our post-processing tools or you can download the results and use third-party tools to make the even better visualizations there are lots of different analysis types and capabilities that you can work with from conduction convection chd conjugated heat transfer will remodel the thermal behavior of the airflow or fluid flow with the solid flow within steady state transit there is a lot of time types of analysis and even more it's not only computational fluid dynamics but also structural mechanics thermodynamics or thermo mechanical simulations all right that was quite a lot of talking now it's time to have a demo before i jump to the platform i would like to first show you the overview of the case that i will be discussing basically we'll focus our attention during this short demo on the baseline case where we have the simulation without the called the division between the cold and hot aisles well the system is configured that but we do not have the insulated aisles to be working um so we have the core cold inlets and the bottoms that are pushing the air downwards into the below the floor the the below the floor then for each and rack we'll have 60 watts of power support supplied through the top plate of uh of the server making it 120 watts for each and every of these slots that you have that you can see on the on the screen okay let's jump to the platform i already opened the project and well whenever you do it yourself uh please be aware that it might require some time for it to load at least when we're talking about this case it's a very complicated geometry and it might need some time for it to be processed but overall the performance this is all running live in the web browser there are many formats that you can actually use and upload right now we are supporting step igs stl solidworks autodesk inventor and rhino we are willing to extend this and we will be extending this base further and that this actually depends on the the requirements coming from our customers there is a question about uh revit there is a rather straightforward workflow that you can use in order to import the tools to uh to sim scale from revit you would save them into another format that is compliant with one of the standards here so either you will get a step or an autodesk inventor uh tool or or other format that later can be imported into sim scale we have customers who do that all the time now after importing them there is one more way to actually get your designs into sim scale you can import them directly from on um shape this is one of the first well this is the first uh online and the best online cad tool available right now on the market if you haven't heard of it you should definitely check it out we have direct integration with with this tool okay once we uploaded the geometry to our platform we will need to follow a few steps that allow us to run the simulation itself first step is performing what's called the domain discretization we will be splitting this room into i would say pixels that will later be used for for the analysis type this is called creating a mesh in this case we are dealing with a mesh that is over 18 million cells we needed to have a sufficient mesh refinement in the areas that are connecting the underfloor compartment and the top part and sufficiently fine mesh around the servers themselves this is also why it takes a second for for the whole system to load i actually have it pre-loaded in the simulation designer so now it will it will be uh loading in the background but we can also see the mesh in this section over here right now i'm visualizing the surfaces but i can also switch to the mesh with wire frame and these are the pixels these are the control volumes that actually will allow us to get a certain precision of the flow inside the room um yes that is it and now once we do that once we load the case once we mesh it we can choose one of the many simulation types that are available on the platform in this particular case we went with what's called the convective heat transfer simulation or a certain turbulence model we'll be working on a steady state problem um and trying to get an understanding of what's going on in the domain now one very powerful aspect of the platform itself is that once you learn a single workflow for a single simulation type you will be later able to mimic it for any other simulation type we're keeping it consistent as much as possible so even a workflow that you will be following for this convective heat transfer you could later apply on a thermomechanical simulation of your server or on any kind of other ventilation problem that you're dealing with and the workflow itself although at the very beginning might seem a little daunting it's rather simple you go from top to the bottom of the simulation tree that appears on the left side in the navigator panel and you start setting up the properties put up gravity make it actually work add a material we're supposed to work with air the air itself will have certain properties we'll need to initialize our simulation say what was the pressure inside our device was there any temperature what was the temperature at the very beginning what are the turbulent properties and so on and so forth then we jump to the most important part that is the definition of the boundary conditions that will allow us to understand how does the system operate itself so for example we'll set up adiabatic so perfectly insulated walls the external walls on the side we'll say that the floor behaves in a certain way we will put the tilts that get the air go in and this takes a second because there's all over 1800 faces that put the air circulating in this system we will have some other walls and then the interesting part that are the inlets in this case we're setting up a certain velocity inlet and we're pushing a fixed temperature we're assuming that our air conditioning units are operating in a certain with a certain precision for the outlets we're using an open outlet so the flow will adjust itself based on the overall flow patterns within the domain then for the racks this is just the walls of the racks we are setting them to be again adiabatic so we don't want to have thermal contribution of these elements to the system because they're actually submerged in this path and then we are applying some other server sides these are the small faces on the sides we are considering that they are not actually generating heat but it is the server top and bottom faces that will be generating 60 watts of power so in this case when you're looking at a server rack you see there is this is the the element itself this is the top rack side and this is the bottom rack side this is where the heat will be generated now there is an interesting problem that we have to deal with how do we make these servers move the air around this domain because in general actually right now there is nothing pushing the air except for the inlet itself that is moving with let's double check with a quarter meter per second that is not much well there is a very smart way to mimic the presence of the the fans that will be positioned between the racks what you can do is create first you can create what's called a geometry primitive that is a cartesian box this is a virtual object that actually does not exist in the domain itself this is a big big thing so it takes a while for it to load let me switch back to the surface mode it will be working a little faster so notice that i created a very small a very small box and this box is assigned to be a momentum source that will be actually pushing the air in the y directions with 10 centimeters per second so any air that arrives here will be pushed still i know what's going on it's actually selecting the surface so it's selecting the whole domain so it will be pushing the air as if these boxes were representing some kind of virtual fan that is moving there is a very interesting question for the crac units are we using atmospheric pressure or any other pressure this is actually uh dependent you can actually use the the values that you have measured or that your supplier gives you that this will be the pressure at this uh at the device itself an interesting thing is that for this kind of simulation you have a relative pressure that you're covering it's not the absolute pressure so it will be automatically adjusted to to the outlets themselves in general you could also consider using suction outlets that would remove a fixed amount of a fixed amount of uh air and that they would be sucking so what is depending on on the speed of the fan well in this case we simply made an assumption that this will be 10 centimeters per second and this is true for all the for all the racks if we were to show all the other cartesian boxes like they are actually present within each rack but you can specify them individually for every kind of device that you have so if you have a specification of your fan which you should have you can decide that well the fan is operating at a much faster speed and use this as a as your reference uh speed for for the analysis in this case as i said for each rack we are giving 10 centimeters per second of the flow and we're simply creating what's called momentum source next we move forward we do not um we reduced a little bit some of the properties uh most of the time we're trying to update these numerical settings so that you don't have to tackle with them and that they will give you some reasonable values as you go then here you have another example of the power of the platform itself since you are able to decide on how big of a machine do you want your analysis to run simscale gives you access from one to 32 cores of computing power for a free account and if you get a paid account or uh you you might also get 96 cores um so and you can run any number of simulations at the same time on uh any number of machine on on any combination of these machines so you could i it's very common that they have customers who are on five ten simulations at once they come back after a few hours and they get their results they are able to process them and work with uh with all the data without having to wait weeks sometimes to analyze and run the simulation itself all right we're almost at the bottom of the navigator meaning that we're almost ready to set up our first simulation sim scale uh one more thing to do is to get some average data for example getting the outflow averages uh some data over particular cell server elements that we could gather and monitor throughout the simulation after all the setup is done all we have to do is click create a new simulation now it will take a sec give it a name and click start i will not do that since i already performed this calculation uh and we'll jump to the post processor to take a look at the results so the simulation gives you an overview of how is it performing based on the residuals you can also monitor the output of temperature profiles the velocities of the result control items as they are developing throughout the simulation itself and obviously you can also show your results in the flow this is a very simple visualization that i'm doing here right now which shows you a threshold of excuse me of temperature but of course since this is a pretty big analysis type we needed to make some post-processing beforehand and this is what i'll jump back into our presentation let's take a look at the baseline case and the performance of what's going on this is a slice at around the middle of the rack and you can see the velocity profiles and the streamlines that are developing within the system the cold aisles are the places where we have most of the flow and the fastest flow present because well we are simply pushing the air through them if you take a look from the top you can clearly see the flow velocity profiles there's a lot of velocity going on around and there are certainly some recirculation zones which are marked over here with this red dashed zones so so our hot air exiting from the service is returning back to the hot the cold side reducing the efficiency of our data center this is an interesting uh section plane that will present that will show you actually this is the thing that is happening if you look from the top you can see as there are recirculation cells which reduce our efficiency furthermore as we have seen in the simulation setup our fans were not positioned throughout all the racks and there was some empty space on the top of every every section this allows for the air for the hot air to recirculate back and go to the hot from hot aisle to the cold one this is what i'm trying to show right now maybe let me grab the pointer once again the spotlight so we are for example here it's very clearly we can see on the second from the right rack that there is a very significant amount of flow that is moving backwards instead of going directly into our into our computer computer air conditioning unit this is another slice this time we are showing the temperature profiles we see the temperature rise at the server sides well we are pretty low when it comes to to the slice and the temperatures are more or less homogeneous because we have a very good circulation of hair inside the domain once again if we take a closer look at the temperature profiles it is clear to see that that some servers will be operating above the desired 27 degrees and even more the red zones which indicate the the hot aisles these are supposed to be the hot aisles they are a little hotter but at the same time in the cold sections we have a lot of area a little above the ground where we do not comply with the standards and where our service would not be properly cooled this is another section this time we are looking at the temperature profile very close to the floor so that we confirm that there is some circulation going on and that we are actually trying to input a certain amount of hot air but as we move upwards these effects are mitigated and the circulation and the recirculation takes over reducing the efficiency of our data center for an improved design this is another simulation that we set up i will not be jumping to the setup itself just to save time we'll consider the improved results and how did the effect if we take a look at the velocity profiles this time there are barriers that will separate the hot aisles from most of the elements and if you take a look at the velocity contour you will notice the effect that i mentioned at the very beginning where we had the three different scenarios so the bottom two bottom rows of servers they are have a completely isolated hot aisle and there is a very nice flow that goes from the cold aisles inside and it's not being distributed on the other hand if we look at the third and fourth row of servers we had two racks that were empty and did not have any servers in them or for example they had a not working service with a significant space allowed this will imply will allow significant air recirculation to happen from the hot aisle outwards to the cold aisle between the second and the third rack if you look from the bottom and will significantly reduce the performance of the door of our system a very similar thing is happening with the last right the fifth one on the top where the air is being recirculated in this very nice circle pattern on the top left you can see a high velocity profile relatively high velocity circulation and this is very much confirmed when you take a look at the slice from the side of course you could have more of this section views to have a better understanding of how the flow is moving inside the domain and then another visualization this time we're looking at the temperature slices very close to the floor sorry this is and middle and this is close to the floor another section with some temperature contours to confirm that we are actually having the hot air being pushed away through the hot aisle and how does it differ from from the one that we are having at the cold ones and again this time we can see that well that we definitely improved the performance of the whole server room but there is more to be done in order to comply with the standard and make sure that all the servers from the bottom to the top are served with proper air quality or proper air temperature these are some more section views uh looking at the temperature profiles we're moving upwards this is in the middle of the of the section of the sections of all the servers and close to the outlets this is at the top where we have the openings which allow which have even more temperature now let's compare the results and see if we actually did change something if you would take a look at the average rack temperature for every row of servers we can see a significant improve improvement in the the temperature we have reduced the temperature at which the servers are operating or what kind of temperature of the surface that we have it depends of course where the rack is positioned and what is which row of uh of service are we talking about but the improvement is definitely that and if we were to use the power consumption or calculate estimate the power consumption based on the temperature the operating temperature of the given rack um based on some manufacturer's uh references that we have available we can definitely we can also quantify the amount of power that we will be able to save and how much or basically how much money we save for thanks to these improvements now i know what you're asking how can i start to go and work with simscale all you have to do is log in to our web page that is simscale.com you can start simulating within a few minutes create by creating an account that is completely for free or you can schedule a demo with us talk with our representatives and they will be very very happy to answer your questions answer queries and help you on your way in getting your simulations working all right thank you very much for attending this webinar i right we are not ending right now i know that there are some pending questions which i will be answering in a second but for those of you who are not interested thank you very much for being here with us i hope that you enjoyed this webinar and i will later on for those of you who had any trouble with the connection or with the audio we'll be providing recordings of this webinar all right let's go to the q a session my favorite part still have some time left okay so uh let's start from the top because i see there are some trouble with the sound there's a question about the the way to apply power to the racks what do i refer to when on top of the bottom i think i already had the chance to to explain that throughout the talk that's why we are leaving the q a session at the very end so so you don't have to uh say jump from one subject to the other so once again uh when we are talking about applying power we were having the top side of the server and the bottom side that would be the top and bottom of the elements so let me just grab another view because this is so this is the top side of the server and bottom side of the server now notice that whenever this is an aside information that i would like to give you whenever you will be trying to apply your own boundary conditions and your own your own simulations whenever you're applying a wall boundary condition with a certain amount of power in this case it's 60 watt this power will be applied on every surface that is selected so in this case we have 912 faces selected if you multiply it by 60 this is the total power that is applied in this domain if you don't like it you can also use the heat flux condition which defines the amount of watts per square meter so that we have a certain uniform distribution of energy that is introduced into the system okay how can i define interfaces between different geometries i'm not exactly sure what you mean by interfaces but basically right now you could actually take results from one simulation at a certain slice and apply them as boundary conditions this is a key word as boundary conditions on the other fluid flow simulations for structural analysis this is a little more tricky and it's not supported uh well it is supported to a limited uh to a limited application range so yes you you can use the results from one simulation as a boundary condition for the others but when it comes to modifications of geometries per se we do not focus our attention on geometry operations we do what we do best that is simulating and we leave the cad modifications and mod and cad operations to the experts like for example our partners in on shape or any other vendor that is providing reliable cat tools uh there was a question about the crac units and the outlet pressure i think i answered this throughout the during the presentation and also how do we know the speed of the fan so there is a question like that's there's a very good note that server fans can be having variable speeds uh working faster or slower depending on the performance of the device so in this particular case we are using a steady state simulation so overall we are considering an operational condition that is that the servers themselves were working for infinite amount of time sufficiently long time to establish a certain flow regime um so in this case you would run the first simulation with the reasonable conditions that you assume that would be working for you and then you could adjust the fan speeds uh for your next simulation in order to have a different performance um a functionality like having a variable fan speed based on the flow this is this is a very complex logic that would require some automated simulation tools which we do not support yet but a very interesting idea i would say okay are there any training videos for students you should definitely go to youtube and search for simscale we have a ton of materials available for free and even more if you're interested we have it on our web page so all you have to do is go to simscale.com let me try to go to simscale.com and go to learning there there are tutorials webinars there's the whole academy there is the forum which is a great place to learn do go there and become even better uh simulation with simulations any are there any free license for students simscale is free for everyone you can actually start simulating completely for free the only catch is that your knowledge you need to share your knowledge with others we really believe in the community we really really believe in collaboration of engineers in order to make the world a better place so if you are willing to share your expertise to create public projects then you can simulate for free on simscale and look at this these are all simulations done by people who wanted to share their knowledge you can use them as a starting point for your own analysis then if you need privacy you can contact us and you will be able to create private projects in fact this data center simulation notice that there is a little lock over here this means that you will not be able to see this particular simulation this copy because we don't want to flood the whole place with 50 different copies of the same project but yeah so go on check check simscale another question have you modeled boundary force so you don't need artificial momentum sources well in this case we are creating uh well what you could actually do for for the whole simulation that we're dealing up with here you could create a little fan over here and you could create what's called an arbitrary mesh interface simulation ami as a mrf multi-reference frame or arbitrary measure interface which would have a rotating domain that is actually a fan but the approach would require you to have a sufficiently fine mesh of the problem and imagine this there are few hundreds of fans over here if you want to have sufficiently fine mesh and sufficiently fine resolution for each and every fan that is present in this domain that will create an enormous system to calculate i don't think that you well you could run them on a supercomputer that would consist of few thousand cores but does it really is it really worth it in this case you are making an engineering assumption that will allow you to have a sufficiently reasonable approximation of the performance of a device and still you get valuable data so it's a trade-off you could also uh argue that you could make a simulation of a single section of of such a such a rack and see how does it perform and then use this knowledge in order to improve the performance of individual fans in the simulation all right do we have a library of items for more realistic simulations for example specific server or cooling units so in this case you would have to look for the public project for the public projects that are here but there are some electronic cooling simulations the whole thing is that you do not put for a simulation like this you are not putting like a block that later will be behaving in a certain way you need to have an understanding of how does it perform and you're simulating the behavior of the whole system so i hope that this uh this gets to the question uh so are there uh any available libraries with commonly used equipment against servers uh crack crack or other devices so for this i have to say no there there are no no direct libraries but you could up you could use these conditions that if you have them noted down then you will say okay this is a mini this is an outlet an interesting topic i'll definitely bring this up to our development team and hopefully at some time we will have a library of such items how can we simulate side cooled equipment like routers rotors well in this case i would say that well the whole setup that we have here is a very uh is a very particular application but you could see some in the public public domain you have electronic cooling simulations of particular devices this is a very nice simulation chd simulation where we're actually analyzing the performance the heat performance of a device itself and later based on the knowledge that you get from a simulation simulation of such a device an individual device you can apply this on a larger scale scale analysis it makes certain assumptions about the performance of each rack next question so for the meshing there's uh that uh that the the we're there's a question that uh i should need filling air but uh it's likely that you mesh the equipment racks and such so basically in this case we are only simulate uh meshing the air all these racks that you have here they are actually well hollow inside if you if you were to take a look i'll hide this this is an empty space it does not contain any mesh inside i like to think about the meshing process as if you were pouring molten lava into into a domain and basically looking at the negative of the domain so in this case we made certain simplifications to to the design of of the server itself as you can clearly see and then we mesh the air domain okay does on shape have programmed racks and cracks that i don't know i'm really sorry you would have to ask the the on shape guys but i know that there they have a very big library of standard elements and even if they do not have such equipment they also have a huge public library of projects and you can also contact other designers who who have them and some of them provided for free some they i think they even have a marketplace right now um there are some questions about buoyancy forces buoyancy forces are being simulated within this project so we are actually having hot air moving up this is why we have to specify the direction of gravity a very common mistake uh well you will get a warning if you do not set gravity properly uh but so back in the day when the warning was not there i cannot tell you how many times i was simulating uh buoyancy flows on international space station simply because there was no gravity so yes we're we're having buoyancy forces in action uh back to the another question about the fans uh for the momentum sources there's only constant velocity something that could change the velocity based on the static pressure difference so in this case you would have to use that there are special boundary condition types that allow for that but unfortunately if you were to simulate the presence of the fan itself as a boundary condition you would have to consider the fan as a black box so there will be a suction outlet and a velocity inlet for example the problem with that is that you cannot maintain continuity of the flow so if we have a hot flow on this side and we're pushing it through the fan itself if you're sucking the air out it's going to the void and then it's being introduced with a different velocity so this kind of joint inlet and outlet sections they are not supported yet and at the moment you would need to have you would need to use the the momentum source approach okay all right can we save a present objects um i'm not sure what you're referring to but if you mean like the say you create a crack and then program it and then save it as a function block to be reused in the future oh this this is the thing so basically you could use the simulation itself uh but uh whenever you change the geometry in a simulation process whenever you're making any kind of simulation you will need to generate the mesh itself so for any change significant change in the geometry you need to rematch the whole domain and this is a significant uh downside of of the whole process at the same time it's it's a necessary step in order to have the the possibility of simulating the void the air flow at least with the methodology that we are applying so uh there is no possibility to actually have the crac as a as a functional block but at the same time what you could do is have it analyzed and then use it as a black box in a way as a black box where you would assign a certain boundary conditions to to a given unit so i would select uh the the section inlet i would select the section outlet i would select the sides and i would know what are the properties in general of such a device so in this sense yes but it's not a single click thing okay uh very good question can i show which kind of simulation type is used in this simulation it's very easy you go to the name of the simulation and then you can find the analysis type in this case we are dealing with convective heat transfer simulation and these are the standard settings for for a problem like this so we are using the k omega sst tool glance model this is a steady state simulation and we are not using businesco approximation we are actually expecting a high temperature difference and the model the simplified model would not be sufficient another question can we simulate the crac failure and monitor changes over time this would depend on the definition of failure but yes you could for example simulate thermal performance or overheating of certain elements and see how they how do they develop over time or what are the conditions at which the device performs and works in a stable way so yes and no but it really depends on on the application that you're thinking about i would highly encourage you to contact our representatives present your case what is it that you would like to simulate and or if you're interested for for hobby perspective post a question on the forum there's a lot of people who would be very happy to answer you and help you with the problems that you're dealing with because like you know it's hard for me to to put actually the the the to have the perspective on the complexity of the problem because there are aspects that we can deal with there are aspects that are beyond the scope of the tool that you have here in the web browser there's quite a lot you can do all right we are out of the question out of questions and even better we are out of time so just on time once again thank you very much for attending this webinar and this demo of simscale platform i do hope that you enjoyed this as much as i did it was a pleasure and i hope to see you on the platform pretty soon working with your simulations and your problems have a great evening have a great day aros osnovsky signing out bye bye [01:03:19] Speaker ?: Thank you.