Exchanges on Exchangers, episode 7 - Heat Recovery and Data Centers Integration
good morning and welcome to the seventh episode of exchanges on exchangers i am kristy smith and i'm cosimo pecchioli today we are very excited to talk with jaime Comella from cloud and heat and birger ober from Valtenfall so can we start off with you're both introducing yourself and telling us about your company hi mate you want to go first all right perfect yeah my name is jaime comella and I work for cloud and heat technologies i'm located actually in madrid in spain but our company is from from germany from dresden our company was started around 10 11 years ago and focused on the heat reuse of data centers and distributed cloud services so this is if we are very very fast in defining what we do is uh are the most uh two important things just an extra comment about that we also use liquid cooling because we think is the best way to uh make this sustainable and to make heat reviews possible hey first of all thank you thanks for having me, my name is birger ober and i work for for Vattenfall. Vattenfall is one of the the bigger european energy utilities and we have a bit broader business so we are we're producing and and distributing electricity and also heat also in energy trading we have a variety of businesses in we're one of the larger wind park developers in in europe we operate one of the largest district heating networks in berlin uh but we're also active in the netherlands and in sweden where our home base is so it's a it's a swedish-owned company very very interesting uh jaime i actually i have a question for you right away um there's a lot of different flavors of liquid cooling uh there's there's cold plates someone says there's a rear door there's immersion chassis base what kind of if you have one let's say the solution which one do you use normally good uh question cosimo actually when we're talking about liquid cooling we must understand the cold plate as a liquid cooling version we use of course in our data centers are there are mixture of different systems uh not never 100 can be liquid cooled so there's always dissipations in the air which are air-cooled actually have to be air-cooled and this is almost always in our case indirectly with liquid cool i mean to side coolers we are doing heat exchangers top coolers whatever you can have there so this is a but of course immersion is another option uh i know you know them well uh this couldn't be also a problem to be integrated into their centers because at the end of all we're talking about pipes water kilowatts power it's always at the end of what is the same and last question which actually it's a it's it's a segue for my question for ober what temperature of the water you you you have in your system yeah thank you all for the question because uh before i was trying to be compact and define everything in one minute but absolutely we we started our company 10 years ago i said before by putting uh black boxes which were ecological heaters in the sellers of private homes the the the people were living there were buying these ecological heaters uh but there were actually racks racks with services inside those servers were used by clients or the people other other companies miles away from that places from those places and the idea of using those servers for the heating system of those houses was because we were able to get a very high temperature from them with direct liquid cooling which what is the high temperature i don't know the Fahrenheit sorry for the American spine 60 degrees Celsius why 60 degrees because with 60 degrees we were able to also deliver domestic hot water which needs to be hot enough to kill the legionella so ober this actually this this comes to you um as as a utility company uh do you think that you have a good use for a 60 degrees water stream um what what is a good use that's that's like that's what you what you need to define of course we have a variety of applications like the first applications that that heimi just explained we have also a business that is uh that works for and with multi-family homes for instance so we have also a branch that delivers heating systems for for especially those homes that kind just explained or described and for these of course the 60 degrees is is fine but i also mentioned our very big berlin district heating system and for that it is really a a challenge especially when you think about raising the temperatures because these are run at above 130 degrees c especially in winter times this varies a bit over over the year of course in summer you don't need that much of uh or that high of a temperature but when it gets really cold during winter times you actually do need this temperature because berlin is three and a half million people and for three and a half million people need quite a bit of heat and you need to transport that you via via hot water and when you have the the respective volumes you just need the temperature in order to limit the amount of water and that's just pure physics uh you have to limit your you have to limit the amount of water you're putting through it so you have to increase the amount of energy you transport per liter or cubic meter of water that's why we need a bit of a higher temperature and and and that of course is technically feasible there are systems around to also push the temperature up uh to that uh to that level that that that we could use it but this is technically and economically net not easy and that is uh actually something that we that we do need uh to work on but at the same time looking at the amount of heat that data centers are putting out nowadays there is just in my view this one single uh senseful or or feasible heat sink actually because thinking about installations i mean i i in my view we're getting news from the Frankfurt area more or less on a weekly basis uh that that hundred hundred fifty hundred sixty megawatt data centers are being announced and putting this amount of heat or electricity then turned into heat to good use is a challenge and there are not many other heat sinks around that may actually make sense uh on on a rather low temperature level apart from district heating systems so i think it's really really important to close this gap which brings actually to another challenge and and high me since you have been around for a few years probably you have experienced that um district heating or let's say domestic use is highly seasonal yeah so how what do you do in summer when you are actually that heat sink that that Ober was mentioning i mean it goes down yeah that's a good question um actually is a matter of dimensioning uh your systems at the end of all if you have a 10 megawatts of install power in data center delivering i would say constantly 10 megawatts you have to cover with this 10 megawatts the demand of uh i would say such a large population that is always consuming that that 10 megawatts alternatively you can always use the the the waste heat for other for the things i mean you can also use in summer for making cooling i mean it's of course thermodynamically thermodynamically not so efficient as compression cooling but if you bear in mind that you're using a garbage and you're using something that you want to get rid of it's always good to use that garbage for making something valuable so heat for heating is the most automatic thing we everybody think about heat for cooling can be done heat for desalination of seawater is also possible green houses in the north when it's very cold as well olympic pools etc also industrial needs for for sorry for heating like uh wood drying this is something very typical as when the nordics in sweden know where they are drawing wood to sell it uh farther so we we might have to look farther i mean in the other applications besides uh the i would say the the easiest one even if it's not easy as burger just said uh to be able to integrate all these data centers all around the world not only in the code of coolest regions and if i if i may add i think that's a very very important point integration i think that's really a key word in this we see it in for the district heating part if you look at what what we're doing in this city is actually we're producing electricity because we also need it in the city we're producing heat because we need it and the same interfaces you have with the data center just the other way around of course we are producing electricity data centers consume electricity we need heat the data center is producing heat so the interfaces sort of are the same and that's why i always advocate for for a bit of a different thinking that that the data center is becoming or will become an integral part of an energy system and the urban energy system around the district heating system is just one example of a energy system uh when you think about the nordic installation that frame i just mentioned of course you have these massive 300 megawatt installations out in the out in the nordics where very little people compared to the amount of energy live and if you just don't have the discrete heating system you will not persuade uh a quarter of a million people to move there just because there's gp it just won't happen so you have to you have to think or to to emphasize on the on the system thinking around it and the integration and and bring these industries together because now where we came from is very isolated business cases and people are very um are not very eager to uh to en engage really in in cross branch discussions because they have been having super successful businesses this goes for for data center operation this goes for energy utilities this goods goes for the wood industry and everything has been worked on isolated very very successful with very very little incentive really for uh collaboration and optimization but it always came at a price and this price is increased or decreased efficiency or not uh not utilized possible synergies and i think that is something that we can overcome at least a certain extent via integration but what i think is uh integration is a super important word in this context super interesting uh conversation actually and and um we have been thinking along the same lines for for a while um but it seems that when a company picks a new location for a data center um the possibility of integration with the existing let's say system or ecosystem of reality is not high in in their list of priorities cost of energy water availability real estate infrastructure of course first because if you don't if you don't have connectivity or power you don't have any chance but i mean the possibility to reuse whatever waste energy you have i don't think it's very high in their priorities right now at least in the us it it seems that that that europe is a little uh ahead of us when it comes to this concept but i i i maybe you can spend a few words how do you incentivize this integration jaime and and auburn were the two of you yes actually in europe there is this uh co2 emissions market right and uh we we sell rights of emission and between industries and so on um europe also has very strong objectives of the carbonizing the lions the industries and so on at some certain point of time i am pretty sure that everything will be linked to the co2 emissions rights and this will be marketed in a very broader way so there is no way to escape that and so there is no way to under the best way of of reducing the the carbon emissions of data center is the waste heat reused definitely because every kilowatt every energy unit power unit is going inside for data center is for producing some some real value of services digital services but if you are converting that into a secondary value and you're converting that a hundred percent into into something that you can reuse then your co2 emissions are are zero i mean so this is right yeah but so ober and and from from your point of view how do you incentivize this integration because currently again i mean i'm talking mainly in the u.s and i understand that europe maybe it's a little more advanced but there's not much yeah as in because because the stick is regulations and no one likes real regulations that what's the carrot as in as an as an engineer it it it causes physical pain to me uh in order to say this or to say the following uh it is mainly about money the rest is just technology and that's really causing physical pain as an engineer saying that it's just technology the rest is about money uh the alternative just needs to be more expensive uh as much as it as i know that people will will cry out yeah you're hurting economy and you're you're trying to kill projects and so on so forth so forth but after all projects will happen as long as they uh result in a in a positive uh in a positive business case uh and if you have a desired outcome which is lower co2 emissions and you are not able to put a price tag on these everybody everybody will choose the cheaper or more economic way and you can't really blame the people or the project managers for for doing that after all they have their own set of incentives and they have people they that are financing uh the projects and it's just a matter of you know it's i don't want to turn it it's a political decision or discussion but after all we are living in there or working and living in an economic and capitalist uh world and and people trying to maximize uh the the the outcome from of what they have so they are after after profit after all and they need to compete with also with with other companies and if there is a choice somebody will take the cheaper choice and as long as they are in a in a competitive environment it's just natural that they do that and you get you can't blame them for that so that's why i think that the incentive scheme needs to be created and i believe this is a via a price scheme and raymond just mentioned it and i think that's that's true so we see it about around co2 prices in in in europe and you see already now what this does to the energy supply actually because all the fossil fuels are becoming more and more expensive we have seen a i think a factor of four uh so the prices for co2 emissions have quadrupled within this year already and and and that is that that creates an incentive on the energy supply and that's that the input factor but at the same time you of course could uh could think about uh just as co2 is in a mission uh heat is also an emission it's being put in the in the atmosphere and you you see this from the uh discussions in singapore i believe uh where they where they see higher temperatures within the cities and they at least partly link these to the high density of data centers there so it is an emission just like diesel emissions or dust emissions and they are limited by law for a good reason and to me heat is moving in the same direction so and if you treat it as as what it is it is an emission then it needs to have a price tag so regulation essentially what you are saying is regulation basically basically yes and or yeah financial incentive regulation whatever you want to call it but yeah it's true when we talk about heat regeneration or hit reuse we take for granted that we are talking about some sort some form of liquid cooling right uh that does that is it possible to think about a heat regeneration of an air cooled data center yes it is possible the temperature level will be lower or your overall efficiency will be lower because you were using the quantization part of the of the cooling machine so you will have a lower efficiency but it's always possible and there are use cases in germany there are a couple of them there is a really flagship project right now in frankfurt from telehouse making some publicity here for them and uh they are connecting to actually a district heating uh system in in the city of frankfort in the northern part of germany there's always a usage of waste heat for elegant farm for an algae production growth and in norway as well you have different examples of that but it is maximized to liquid cooling and this is um physics and of course liquid cooling we believe in liquid cooling because uh you are much more efficient you go down in your in your opex so in your operational cost you go down and you go up in the heat capture you you can you can have and in the in the quality of that of that heat you can store it much better you can transport it much better uh at the end of all uh is is the medium isn't to to use sure so we definitely see that the future is going towards the liquid cooling but maybe not as quickly people are racing the liquid cooling yes so what do you think this has been the hot topic out there what do you think is one of the biggest hurdles against um using liquid cooling price okay uh totally uh again price why because of what of the hardware mostly i mean because if you're thinking about the data center the 100 kilowatts uh air cooled versus 100 kilowatts mostly water cooled you will install less chillers so you will actually reduce the capex for that same data center right if you're cooling with liquid cooling with high temperatures of course if you are needing low temperatures then it should be very similar because you need to cool down the the water for that and then in the opex you are going down because you are using much less power your pue values are going down and so on you can even sell the heat so you get to go even down actually artificially you're putting down your your power price with that but in the data center 60 50 to 70 percent of the cost are the hardware cost and hardware costs for water cooling for liquid cooling are still too high they are still not so massively produced as the other ones this is of course one of the reasons and uh and but this should change i mean we we do also adapt hardware that is air-cooled into the liquid cooling and actually it costs very very few money to make the cooling blocks to adapt them inside i mean we have to get rid of all the fans that they come from the from the factory i mean if they came from the factory without that it will be even cheaper and we can we assume cost from from five percent to ten percent more i mean in the adaptation of of of air cooling service into the water cooling servers i mean but uh the the realities that the industry is still not so mature for that uh mentality is in my opinion not a problem because we do use put water and technology together already in the data centers with side coolers top coolers every other day so this is not an issue i would say i mean it's more for me it's more the price but it's solvable again a challenge that that data centers are facing right now uh particularly in the us and particularly some areas that are popular with data centers that are very dry and so water consumption is becoming really an issue and one possible solution as jaime was saying is definitely liquid cooling but so and one resources now has been seen as as critical already which is water and communities are pushing back installation of of new data centers because of water consumption if you you and envision a an environment in where data centers are kind of embedded in urban realities and cds and so you can use do you ever see actually power availability becoming an issue uh when you put together all these industries in one spot um it's not really about at least uh in in europe it's uh if you want to move right now power can be a challenge of course but it's not the overall availability it's more the time frame is be difficult so of course you can have you need always the three things for a data center project right you need the you need the site you need the power you need the connectivity and everything else will will start will sort out but if you want to move right now it could be that the local power grid is just not fit for another very very big electricity consumer so at least that's what i know from from most of the german market is that if you apply for a power connection uh early enough you will get the power because there's no overall power shortage of that because europe has a rather strong electricity grid and the electricity will arrive at the desired punch just a matter of time so electricity as such is not right now it could be a challenge yes if you plan the normal a few years ahead then normally you should run into a problem and because there's a there's a big attention let's say on what what kind of energy it's used for data centers so are you talking about traditionally fossil fuel generated power or renewables that that's actually really interesting uh interesting discussion especially when you think about the discussion about uh green washing certificates for electricity and so on and so forth i think this while this discussion sounds super easy in in at first glance i think i personally think it deserves a bit more of a uh differentiated discussion because uh if you just take two scenarios and we touched upon these earlier in the discussion uh take the megawatt data center in the frankfurt area so you already have a scarcity of land uh and but you have at the same time you have the district heating system which could serve at least theoretically as as your as your heatsink but since you already are in a position where you don't have uh you don't have land readily available you you don't have the ability or the possibility to directly supply it with the big solar farm or with big big big big windmills uh let alone hydropower i mean um this you you can't plan for this so and there's your dilemma put it in the in this in the next to the city of frankfurt where you probably need the capacity because it's latency relevant for for all the banking that's that's going on you just can't put it somewhere somewhere else it needs to be right there and at the same time you have the possibility to reuse the heat but then you can have renewable direct or direct renewable supply so is then putting certificates on top of that a bad thing i wouldn't say so because you have done the best that you can and the at least you have you've done the um you have done the um the heat reutilization but the other example that we had is like the 300 megawatt installation in the north of sweden uh renewable electricity there i mean there's hydro abundant availability of hydropower and that's then not a problem at all but i don't think one is the better or worse system it's just a different application and the different set of uh boundary conditions actually i think the only bad thing you can you can do is just ignore all of these resources and the possibilities that you have and part of my ignorance here of course i i'm not a very utility expert but can you just move the energy from a solar farm or a wind farm to the data center with a with a new power line it's that's normally or there are different steps for supplying green power of course you can have origins of certificate certificates of of origin or you're actually matching directly the supply or production of a renewable energy system to a consumption of course you you can do that and that and the discussion is really ongoing right now is this really green electricity or is this still i mean you can't control the electron where it flows right so it's just the electron goes wherever it's the it's it sees the least resistance so physically you cannot rule out that this is coming from a coal plant from somewhere in in in europe right if you if you buy 100 megawatts of solar power power for your data center someone else's and and 50 percent of that comes from a a natural gas power plant someone else is not using that 50 megawatts right exactly this is then balancing and that's a big question if if that that is okay uh because at the same time it's it's a balance right uh and if if you or if if if the data center operator buys this amount of electricity produced by the wind farm or by the by the solar by the solar farm it's gone right and so somebody else cannot use uh it cannot use renewable electricity and i think that's a bit of the core of the of the discussion around that because uh normally the the data center operator is is in this rather strong financial position so if they basically buy what's being produced from renewable electricity just to be able to say that they are green it's it's it's not a pure economic it's also a bit of a not philosophic ethical yeah something in that in that direction right so it basically you push you push people that don't have really the free choice to use non-elected non-renewable electricity uh because there's just not enough being produced in that sense and then i think that's a bit of a discussion that's why why why it tends to be called green washing i don't see it that black and white uh personally i believe whenever it's it's it's physically possible and and doable i think it should be attempted and done but also see very realistically that there are limits to this demand super interesting conversation we could talk for hours but unfortunately we have to wrap this up but yeah i mean this is i i really think that the location integration uh hit reuse and and and again integration with the industrial ecosystem let's say uh should be one of the highest priorities but certainly it's it seems that right now it's it is not and and cosimo something we haven't talked about uh is about seeing the whole data center ecosystem as a connected ecosystem you know at the end of all you have plenty of data centers interconnected and nowadays the typical setup is to build data centers that are totally they can work isolated they can they have to have a lot of redundancies a lot of machines involved in the data centers just to be available of having 99.9.999 percent of the availability per year while this could be actually
ensured through virtualization at the end of all you can you can distribute data centers actually is something we are always it's our vision as part of our vision to distribute data centers that communicate each other as if they were the same data center as if they were the same building so i do identify you can do a very smart load balancing between different locations depending on the variables you have also if this doesn't have any any effect in the service you provide to the final client of latency or so on some applications do not need a low latency and are used for maybe learning models machine learning models or so on you can move loads where they are actually optimally used or optimally actually yeah optimally optimally used are located yeah for example if in the northern part of sweden there is need for heating then you can shift the loader an it load so you can create that heating and the little fall is your is in a holistic way uh better and the functioning of the data center and energetically physically way better so this is something uh actually is this is not still there but i'm pretty confident that with the edge revolution which is coming we are actually announcing that since years since years i know that i mean it's like h always h but now with 5g we all hope that this push is coming and that's really uh will produce i would say a development of edge locations small locations where they are needed actually yeah and where where the data is needed for for the close calculation to the place where the the the clients are and there also you can better use this waste heat reviews yeah the edge the edge would be probably another whole discussion because that opens a lot of other opportunities and and challenges i i i agree yeah but yeah excellent point jaime i'd like to to raise one more point or again uh and and also put a perspective on the on the temperature level because we only have touched upon this briefly and so for for for the use of the heat the higher the temperature the exhaust temperature is the easier it really is and changing it from the uh formerly 25 to 28 already now with the new ashrae stand that's the 35 degrees c somewhere around that uh the envelope that is already a bit better uh but being able to receive uh the heat at the temperature level that cloud heat for instance is providing is a real game changer for the for the reuse of heat uh so it's really quite digital at 35 or at 28 degrees c it it normally is not possible to do this in a in a technically an economically uh viable way but but the 60 degree that the cloud need is able to to deliver there is really really a game changer uh and that's also a very very easy reason why we have the pilot project uh that we are doing and just stockholm together um so i need to i would like to stress this that this is really really important for the injuries thank you both so much for being a part of this discussion it was very interesting and i definitely think it's going to lead to many more discussions like like you said we don't have enough time in this short podcast but um be interesting to hear this conversation um continue so again thank you both very much and as usual to our listeners if you have any questions you can find us and our guests on linkedin i am christy smith and i'm calling pacquiao and we hope you enjoyed our podcast and we'll see you next time on exchanges on exchangers
2021-09-05 22:50
