Webinar: Containerization, Cloud, Collaboration: Cutting-Edge Approaches with Altair and CIQ

and night shows as well every day what talking about oh wow look at that little teaser morning recap afternoon recap so join in oh wait okay so you're saying like coming here live like coming on to the YouTube channel live every day twice a day oh my God that's amazing what we're shooting for it's gonna be a lot of effort there's a lot of coordination that has to happen there but these people know who they are thank you cool cool I'll make sure to put my face on every day we'll ready to party yeah we're gonna bring people in hopefully have uh customers come talk

so it's gonna be exciting looking forward to it all right I love it and so you know we were talking about collaboration so that little intro that you did uh you know talking about ciq and collaboration that's actually one of the things that we're going to be talking to today so we have a special guest I mean of course we have ciq special guest Dave God love coming on we have special special guest from alter an amazing man named Eric so I think we should probably bring them on there he is awesome welcome everyone so much hello hey good to see you guys again Zane

Rose wonderful to see you Eric we were just talking a little bit briefly before uh we came on about how it's been too long yeah my pleasure to be here again in your webinar that's that's really great amazing how you you manage that and yes I listen to you talking about SC so I hope to meet you in real life there and this will be also great moment and yeah it's good that you will brast many information also during SC absolutely so real quick Eric why don't you introduce yourself and I'll go to Dave in a minute yes so um myself Eric L I work

in alter since very long time in fact I started on did my my career in in radios so I will describe a little bit after what is radios but basically it's a finite element solver and let's say my my original background is is parallel Computing so my first task in radios was to work on the parallel version MPI was already there at that time when I started and then I continue to evolve in in alter and in radios being in ch of the world radios development team since seven years now and since one year in charge also of open radios the open source version of

radio so I will let you give you more detail about that but really enthusiastic about this new way to do collaboration work and to collaborate with companies like like you very exciting and super happy to be there today excellent thank you Eric Dave I think everybody knows you but why don't you go ahead and introduce yourself again for the 50th time well there might be people who are tuning in for the first time to this uh webinar so Rose introduced me as a special guest I'm not a special guest I'm a total normal guest but I yeah so I'm Dave godlove I uh used to

be a scientist at the NIH and then I became a systems uh administrator staff scientist at biowolf120 from there I became interested in containerization I was kind of an early uh Community member in the uh Singularity Apper um community and um so I've been around uh HPC containers for some time and I've been working with them and uh now I'm a Solutions architect at C ciq continuing to work with uh containers as well as some of the other awesome technology that we're developing and I'm happy to be here talking with you Eric and talking about this this great work and this great collabor op ation

between uh uh ciq and yourself here and you're always special to us Dave well thank you Z I feel the same about you and Rose both thank you I know you do um I also Dave just want to uh congratulate you like we worked at a different company together and you had like moved into management positions where you were doing because you're good at that Sorry Charlie you you're good probably didn't want you saying that out loud I talk to some of the people that I worked with in a management capacity before you say that because I don't know how good I am at that

well you were and I did actually get feedback from your team because that was part of my job is to know what was going on so yes you were good at it not being good at it doesn't mean that you please everybody but you you were okay anyway so side note but I remember you very clearly telling me I would so much rather get back into the command line into the code into writing stuff into creating stuff and putting things together and working with people so I am really excited for you to be part of this partnership um with alter and you know the other

companies that we work with as well so I know that you really kind of specialize in containerization specifically oper um but why don't you tell us a little bit what we're talking about here so it's containerization Cloud collaborations The Cutting Edge approaches with Al so what is your kind of like Viewpoint then of course we want to hear what you think too Eric and hopefully we're all on the same page I mean that's that's the point right but sometimes people perspective so tell us about it okay so yeah you want me to to introduce a little bit the code before or you want Dave to

introduce what you did with the containerization what what you prefer Eric if if it's okay maybe um it would be good if we start with a little bit uh talking about radios itself and the code that goes into that and then we yeah we can move from there yeah yeah great great so yes so just to recall that radios is used in the industry uh to run what we call crash analysis and also impact simulation so in in historically in Auto in AO in Rail and also food structure interactions in defense and civil and um let's say uh it spread to to other domains like

Electronics consumer goods Sports Healthcare um driven by this increased pressure to from our custo customers to differentiate from the competition to innovate faster so what was let's say restricted to few domains now it's it's uh it's more used in in different Industries so for example to look at the behavior of a cell phone when it drops you need a software like radios to look at the product to be sure it's safe it it won't let a break and so on so the the idea is really to to use those type of tools of solver to do what we call virtual prototyping that is to explore

the the behavior of the product at the early stage of the design uh so without having to to build the cost Pro costly prototype and to do a physical test campaign and to go back to the design after you can do everything virtual and really optimize the product by simulation with the idea to to build safer product more sustainable product while optimizing cost and reducing time to Market okay so this is really the the IDE behind using the the simulation and to further democratize the usage of uh this type of simulation about a year ago alter launch open radios the open source version of radios

and the the genis of the of the project was really to coming from the critical challenge faced by the manufacturing sector and especially the transportation industry thinking about this move toward immobility driven by let's say the the need to have more sustainable product to address Environmental challenges related to climate change um so if you if you look at for example how we we build car this let's say increase the challenges to build them and also to simulate think about the the effect of having a battery pack into a car uh and toh to assess the risk of short circuit of theral thermal Runway after a

collision so you need really to to assess correctly and to have all the the capability into the code uh to to to study the material and the effect post crash of such incident so um let's say the the answer from and the conclusion from my team was really that with Legacy code developed 30 years ago we cannot cope with this fast transformation that requires constant adaptation of the software and also specific more and more specific knowledge so more collaborations so so really then this descriptive approach with open radios to to democratize usage of explicit Dynamics and to build this uh this active community of users

and also contributors with Engineers professors students researchers to increase the The Innovation pace and address those industry challenge with state-ofthe-art research with this uh open platform so it's really what we are we want it to achieve and uh maybe you you want to to explain maybe what what you did and how we collaborate because I think that's a wonderful example of how bringing people with different knowledge all together we can do something really great all together I think it's it's wonderful example yeah absolutely I'd love to jump in now so thanks for that background Eric and so I think what I'm hearing you saying is

that um alter decided uh to open radios and to to create the open source uh version uh open radios really to accelerate um the technology and to accelerate the growth of the project and so that's that's kind of exactly you know what happened right off the bat so um shortly after uh alter made the announcement that that open radio was now a thing um some people at ciq um got interested in it and um in particular my colleague yoshiaki Senda uh said okay well now this is an opsource um code but um researchers are still going to face some hurdles trying to install this software

and trying to use it and we can help with that by containerizing it and so he jumped in and you know I think within a matter of like a few days after the the code had been announced he had created uh a container and had written a definition file and um ultimately ended up creating a poll request back to the GitHub repo where the code is so that um you know this this code can now be easily containerized and you know this is I really get excited about this project because um you know what what really gets me excited about uh working with you know

with with this type of technology is when I see real world benefits and you know a lot of times um you know you know that there are real world world benefits when you are working with uh HPC researchers but you don't you don't see them as somebody who is helping to just facilitate their research you know you you see that you got a researcher unstuck you kind of know vaguely that they're working on interesting things um you see that you helped out you know the HPC uh you know organization that you're working with and you got them you know unstuck on something but you don't

see the actual real world benefits and you know when with a collaboration like this it's like well you know wow you can really see how this is helping um a lot of people you know do really important stuff crash simulations helping the safety of people who are driving in automobiles and you know that the safety of people who are using you know electronics and other consumer devices so I'm really excited about this collaboration and this work I think it's it's really you know it's really great um Yoshi uh went further uh than just creating just showing how to containerize this work he's written a really

nice series of blog posts and um has also been talking about uh he created um and you know helped to create a um a pre-installed Oracle VM which is available as one of the offerings on the Oracle Cloud that has open radios and pairview pre-installed so you can just spin up an instance and you know tell it that you want a big beefy instance with lots of CPUs and you know you can go ahead uh going that route once again you don't have to install anything you can just go ahead and spin one of those up and start analyzing your data um and then we've

also been working a little bit um so that gets you to you know cloud and and how do you use this software you know number one in the container anywhere you want to use it number two on a cloud platform on a a big instance but then um how do you so so open radio is a great fit for app tainer because it's HPC technology it's you spoke a little bit about your contribution um making sure that open radio well radios and then open open radios uh can utilize MPI and so how do you how do you use that to its fullest extent now how

do you scale that Beyond a single node and that's that's the next thing that we've been kind of working on too we've been using um this open radio container with our fuzzball platform and scaling this up to not just you know 64 128 uh cores or whatever but actually uh you know multiple different nodes multiple different cores within multiple different nodes and experimenting with that and seeing how we can speed that up too so we we really had a lot of fun with this and hopefully we're really helping the community out a lot as well and enabling people to to um run these models and

do these simulations and you know hopefully helping to get a lot of productive work done yeah that that that's really excellent and uh I mean um maybe I can let's say comment a little bit how it is helpful for the the community because in fact with this approach of containerization it it allows to to automatize source code synchronization from GitHub compilation execution everything in a container so this uh this avoid to search the different components with the risk of version mismatch something like that you have everything and it's becoming like a press button I would say very simple easy to use and in fact for

for open radios is it's critical uh because uh I need to explain a little bit how we architecture between open radius and radius um that's important to understand for the the audience so the the open radius is our let's say Upstream development version uh it is continuously improved by this active Community uh the code is freely available so everyone can download it and use it but and also can contribute to it so do modification of the code uh so this means that open radios is constant ly evolving okay so everyone can access the latest and the greatest contribution from the community okay and use that

to test new feature or to add new features and uh even if we uh put a lot of care to to keep open radios very stable by running automatic regression test by having a pool request reviewed by a professional maintenance and so on it's still a development version so you you take the latest and the greatest but at the same time you can have the regression with feedback we can improve it add new functionality new enhancements so I mean if you work with open radios it's very important to have a let's say smart and efficient way to update it automatically and with the the work

from you C IQ it's it's becoming very easy to do it and then you you can run it uh on anywhere you you install Rocky Linux and container or you can run it in the cloud as you you explain so I think that's that's very important okay I just want to add also that um the way we ar it is that we have the open radios and at the same time we keep also a commercial version uh because we have some other customers that maybe prefer to wait a little bit to have something more stable more qualified to wait that we solidify we improve the

the the the the the new features so they are ready to wa to wait but to have something more stable and with the the support from the the professional alter people with maintenance documentation and everything bundle this is possible uh but at the same time for people that really want to interact to collabor right having this access to open radios and this easy way to compile it on to update it is is very fundamental and at the end this will also increase the The Innovation and the let's say interest and the let's say easy access to the to the source code which is something that

we know that even if for us that develop the code it's easy for people in the community you really need to have things which are really streamlined very automatic very easy to use so that that's great so Eric I I know that you can obviously containerize open radios I'm assuming that means you can also containerize radios and when you do that and you put it in Cloud how does that impact the licensing model you guys have or have you had to change the way that that works because of containerization yeah so so the for open radios I mean there is no license it's open source

so I mean this is a this is really great because this means that you can you can run let's say many simulation with without increasing the cost and if we look for example at new technology like AI machine learning where you really use the simulation to let's say build tons of data and to use those data to improve your your design by generative design and technique like that you need to to run more and more simulation so having let's say the possibility to use open source software without license cost is very important whether it is containerized or not okay uh for the commercial version um

alter is quite unique in in the simulation scape because we offer a business model which is very flexible and which maximize the return of investment of our users because we use the same pool of License to run radios but to run the other uh solers to run the pre and post processing all the catalog of application available from alter you can run it using the those units and you can also you use those units to run let's say partner applications so you see for for alter customers running the commercial version is also almost free okay so I think this is this is really appealing so

I see more the the difference between the open source and the commercial version between people that really want to be participating to the development be involved in the community in the future of the software then the the open radius is the right tool or some people that prefer to have really deep uh let's say qualification of the code and things like that and then the the commercial version is also very important and very stable and for sure what was done with open radius it can be done also with with the commercial version regarding containerization and like that it's really exciting because we we do run

across quite often third parties software that has a license associated with it can become very difficult when you you move into the cloud or you containerize and move into the cloud so it's exciting to see you guys looking at it from a different View and finding a way to make a business model that is flexible enough to do that so thank you guys very exciting and maybe one another aspect which is also very important with the cloud is that uh you have both uh the way to access let's say the software easily but also to have let's say access to Hardware Hardware capacity okay uh

so and this is very important because what I didn't say is that and explain is that radios is using what we call the explicit time integration scheme so this means that uh uh to to run a simulation uh which is very highly nonlinear okay um like car crash like like a drop drop test of cell phone you need explicit time time step solver uh because it's a let's say it's not possible with implicit kind of simulation to converge okay it's it's too too highly nonlinear and uh it's possible to converge with a spe code because you use a very small time step so this means

that you have big models and you need to iterate a huge number of Cycles so something like 100, of Cycles at each cycle you need to solve equations okay to compute stress and Str at each node of each element and a typical model in the crash industry is more than 20 million of elements so you see how comput intensive it is so to to be able to run simulation in a decent amount of time we put a lot of effort to make the code parallel so to be paralyzed with MPI and open MP and a mix of MPI and open MP and having access to

the code means that you can use the latest and greatest CPU so you can scale in by having CPU with a huge number of cores and you can scale out because then you can use in parallel many instances many nodes uh by using MPI and with the with the cloud and with the open source this means that you can do it at a very low cost and this is very interesting because then this means that you can run a simulation faster but you can also multiple simulation in parallel so you can let's say innovate faster at the end okay so that that's what the the

the cloud also allows users I I think we should remember too that some of the viewers uh here on the the webinar might not be familiar uh with this software maybe now would be a good time um to just kind of do I can do a little bit of a just a visual demonstration just to kind of um you know make sure that everybody's kind of on board and um let me go ahead and show so so we've been talking a little bit about um about this uh this open radio software and what it allows you to do and you know essentially it allows you

to model uh as Eric just said at multiple points in time um you know what it looks like for for instance a car to crash and so this is the results of um this is just in a single time stamp uh we're we're reviewing this in pairview but this is just like a single Tim stamp of a particular Toyota Automobile and this is um these are real data from a real simulation that we carried out um as as part of this effort to try to containerize this code and I mean you can see this is this is just a single time step so uh it

took you know multiple different time steps to to get to this single point and you you were talking a little bit about how many different um elements is present in one of these models so I can I can go down here and I think that I can if I can remember how to do it I can make this um a little bit more opaque so you can look inside and you can kind of see you know all the different little elements of all the different pieces of this automobile that all have to go into this simulation and interact with each other and you know rupture

at certain points in time and so on and so when you put all that together um you end up you know I've got something I can show here you know here are multiple different time points and this is something once again that yoshiaki Senda uh put together but you can see the color coding here gives you an idea of where the stresses are being absorbed in this automobile and so you can see multiple different timestamps as this crash goes along you know where all the stress is kind of carried mostly through the hood and the front end and then gets carried through the frame and

here and and so on and so forth so this this is the kind of information which is um you know which you're modeling and so now because of this you don't actually have to you know go and build a car and crash it into a wall and have a bunch of sensors and gather all the data and everything instead of that you can just you know have this uh have this digital car and crash it multiple times and gather all the data you know as you change things and so on and so forth so this is you know I I know that you already explained

all that Eric but I think that sometimes um you know a series of pictures can really help somebody understand a little bit better you know what what's what what the software is doing and and how it works I I I wanted to to talk a little bit we were talking about the open sourcing of this so maybe now would be a good time to ask I'm gonna step on your I'm sorry Zan I'm gonna you're good do it if you don't mind so I like I was curious I went over to the the GitHub repo so you're about we're about a year into to this

open sourcing effort and so I went over to the open radio GitHub H repo and I noce there's a lot of different um contributors and there's a lot of different um there's a lot of issues which have been closed and uh some ones that are still open and I also see that there's other things within the open radios um uh group such as the model exchange which I guess this work facilitates as well because it allows people uh more freedom to be able to to hone and uh create and you know um help help to you know make better the model so I'm wondering can

you comment on how this this open source collaborative effort is going a year end yes it's it's a good uh point because I mean we open source it was on September 888th uh 2022 so about let's say a year ago uh and in indeed uh it's really nice because ciq was one of the first supporter uh and contributors to to this project project um indeed the day of the launch we were surprised because we we launch open ros.org on at the day of launch we add something like 40,000 connections to the to the website so it was amazing to to see how many people from

all over the the world uh coming to to see and to learn about this initiative I think it was also because it was the the first let's say big code like that in it that went open source and since then we have let's say a lot of traffic uh I mean uh I think that we have in average 400 download per month so for kind of still Niche application it's it's a lot I mean compared to the type of software which are here since long it's it's a lot of download okay and we have very let's say um fast growing community so um I mean

we had two events we had the first event it was the last December in in Paris and we had also the the the honor to have CQ participating to this first event and it was let's say a French event with with alter and then it was a nice success and then we we decided to have a musers day uh so this happened in end of June in aren in Germany and this time it was a full day event and again uh very nice to add CQ sponsors the event and also went to do a presentation it was your colleague B adamski that went in person

I think that all the community was there and very exciting about the different presentations by the way all the presentation are available on open ros.org presentation so you will retrieve a presentation from CQ but also for all the let's say the community around radio so we have people dealing with with battery uh we had the University of aan which is working with open radios on the pionering project of skycab so you will see also the explanation about the the sky cab which is a really a nice way to illustrate uh this transformation in the in the in the transportation industry okay with new type of

things where you need to design let's say to deal with composite with batteries with autonomous vehicles something like that so it's it's really it's really important and it shows how Dynamic is the is the community and uh as you said in parallel to those event on this communi that is growing every day we try to put more things available because we think that it's really important to e collaboration and to e access to the software so uh we work on uh let's say having um readers uh to be able to to transform the code and to support for example output with par view that you

demonstrated to have workflow using other tools other open source tools uh we also uh put models available so we we have this model exchange database that is also free models for everyone so everyone can can use the models and start understand how to to develop models with radios I think that's also participate to democratize the usage of explicit simulation to to give models and models also that you can use to build more complex models so the bricks that you can reuse and in a very free way because it's also open source model and yes I think that it's uh it's growing every day and it's

a worldwide community so that that's nice so we we want also to continue and probably we will have a users meeting in other region thinking probably in the in the US next year something like that in order to continue to to have the opportunity of everyone to to join the community and to participate remotely but also in person uh so that's that's very nice and one thing that I didn't say that in addition to all the help from ciq to automatize the code we also want to to give more insight on how to use the code how to interact with preprocessing with postprocessing uh how

to set up models and things like that and how to develop with with open radios so we also launch a YouTube channel so from the website you can retrieve the YouTube channel which is at open radios community and inside this YouTube channel you will retrieve let's say many tuto it's a short video so you can look what you want and get inside about how to join the community and to start work with the uh with this project thank you for that ER and we are posting those links sorry I double up yeah I see said the next one's coming back to the YouTube channel for

open radio so thank you for pointing out aigor um oh lost my place here sorry Dave I don't me step on you if you're gonna ask something Rose I Heard a Voice I was just gonna comment that I love open source open source is awesome it's really great to see the community just come and you know collaborate and just come together around a project the way it does yes Amen on that and so I think that Dave you're going to show us something yes are we about to like share a little Dave screen and or if now's an appropriate time to do so I can

yeah so like we've talked a little bit about the the effort that has gone into um creating uh the the definition file to containerize um to containerize open radio and then we' talked a little bit too about um the effort that uh Yoshi um carried out to create the um the VM on Oracle cloud that is you know pre-installed so you can just grab it and you know scale it up to however big of a single machine you want to do but then the next step is you know obviously um this is one of the reasons that Apper is such a great fit for open

radio is this is MPI enabled code so The Next Step Beyond that is how do you scale that to multiple machines and you know start using that across an entire cluster either on Prem or out in the cloud and so we've been working on that problem a little bit with our um fairly new uh uh with our fairly new product fuzzball and you know I can go through a little bit how we've been working on um containerizing and running uh running this as a as a job with fuzzball running this as a workflow with fuzzball okay let me uh go ahead and share my screen

again so just um you know if you haven't been following or if you kind of forget or whatever so um you know I like to tell the story of fuzzball kind of this way so uh about five or six years ago um people within the HPC Community got really excited about containers but the existing container solutions that were there were not a good fit for HPC and so because of that the open source Community got busy got to work and created a solution a container solution that ultimately became obtainer which was you know well suited for HPC and we've seen the success over the past

five or six years of of that effort and just how obtainers really taken over uh containerization within HPC now we're at a similar inflection point where um the the community is excited about orchestration and we're looking at job orchestration we're looking at container orchestration and we're seeing that the orchestration solutions that currently exist while great for what they do are not a good fit for HPC really for a variety of reasons and I would say that the the biggest kind of underlying mismatch is a focus on Services instead of a focus on jobs so in high performance Computing what you want to you know when

you want to when you want to orchestrate uh containers you want to orchestrate them based on jobs and that's kind of a completely different Philosophy from what it currently exists out there so enter fuzzball so this is this is what fuzzball is for is for orchestrating uh jobs specifically containerized jobs and so I've been working Yoshi and I have been kind of working on this together back and forth across different different time zones and different work dayss but we've been putting together um a uh workflow that will run um the basically the model that I just showed you previously uh the Toyota yurus crash test

that I just showed you a few minutes ago um and you know that's it's a kind of a nice model that that we can use to do this so I'll kind of Step you through and kind of show you what this looks like um so let me go to the maybe it be better to go to the workflow here so really this is a workflow which consists of three different jobs there's a pre which kind of does a little bit of setup there is a workload which actually you know calculates the model and does the heavy lifting and all that and then there's a post

which essentially all this post does just tars up the data and then put a place where you can get it later um so if we look at this there are a couple different volumes there is a volume uh called called input and that volume is an S3 bucket uh where I've actually placed this model and you know some of the files that we need in order to run the model uh maybe I can make this bigger hopefully that messes up the perfect there we go okay so um so yeah so uh there's there's this input S3 bucket and so the this this volume is like

sort of mounted onto the job you can kind of think at runtime and then we've got um another volume let's see which is going to be this uh open radio volume and this is where the the data is untarred to and actually the work takes place and then ultimately where the where the uh final output is tarred up and then moved um so if we go in and we look at these so the pre um portion of this if we look at the environment here um we can see that we're using this open radios container that we have up on uh that we've got up

in the um Google artifact registry and then if we look at the actual command that's run um so here we're essentially just just doing a a un taring the data and then we're CD to the correct directory and then we're running the actual open radios command which in this case is starter linux64 GF and what this does is it sets the model up and it sets the model up according to the number of different ranks that you want ultimately to be in your MP job um in this case we are specifying that we want 72 different ranks to all run simultaneously okay so now if

we go into the workload portion of this um we see the command that runs down here and that's another open radios command uh it's uh engine Linux 64 and this basically you point it to the first file and then it it kind of you know it it uh invokes or is invoked by an MPI command which then looks at the number of files that it has and you know starts to to work on the data according to that now how do we get MPI to work here across multiple different nodes well this is kind of the cool thing about uh fuzzball is that if we

go down here we see that we've selected the job type to be multinode um and then we've said uh that the implementation in this particular case is going to be open MPI and that's because this um this uh image we have compiled with support for open MPI in the image but really all you have to do is you have to click this button that says you know I want this to be multi Noe and I want it to be open MPI instead of M piter gas net and then um excuse me when you go over here resources your resources just have to match uh you

know the core so so in this case I want to go multi- Noe I don't want to just have everything run on a single core so I want 18 different cores per node and uh then if we go back and look the the number of nodes that we're going to do is four nodes so that works out to the 17 different ranks that we need so that's how all that works it's a simple as that and you know if if you want to click around and create your job here that's great um if you don't want to you can go over here and you can

so what under the hood what this actually does for you is it creates a yaml file and you can you can you know everything is specified here in this yaml file that you you can then submit and I find that it's useful to kind of go back and forth between these if I don't really have any idea what the yamel spec ought to look like I'll kind of rough in some values and then I'll you know save it and then it'll generate the appropriate yaml for me and then I can go in here and I can fine-tune things and look through and kind of see

what the values ought to be but if we go down so all these jobs are dependent one upon another so we've got this pre-job here and we've got the post job they're kind of out of order but if we go down to this workload job uh you know and if we go down here to resources we can see once again multi node four nodes 18 cores that's how all that is specified within the Amo file um all this can be so this is all apid driven this can all be carried out uh you know either at the command line or through this guey but you

know once you're ready to go you just start the workflow give it a name or don't give it a name if you don't want to and just you know get it going um and then if you go to the status you can see that the workflow is running um and it's really as simple as that so I'm going to go back to this takes a little while to run so I'm going to go back to the previous workflow and we can kind of like look at it um let's see and I've been tinkering with this so I've got several of them that uh have have

failed or been cancelled but if we go back to this one here um so if you go in here and look for logs uh let's look at all the logs actually let's look at all the logs so this is going to look familiar to you if you've ever run open radio before so it kicks off and says okay I'm going to do this with 72 processes and you know here we go and then the um the data is ultimately you know tarred up in this job here and then at the as the final step we transfer the data back out to this S3 bucket where

you know I can download it to my local computer and use it to visualize like I showed you earlier with paraview or you know do whatever I need to do with it so that's what we've been working on to kind of take open radios to The Next Step which is across multiple different nodes and it's just it's easy to do because the architecture of open radio is so nice and also because you know fuzzball lends itself so nicely to to running containerized jobs with with uh MPI yeah really impressive and I just saw some let's say community exchange on the Forum about MTI noes and

thing like that so I am sure that the community is will really appreciate to use such tool to deploy on smt noes yeah absolutely yeah um and you know if you're so that's actually running uh on the cloud under the hood it was actually running on AWS but you know fuzzball runs uh multiple different places you can do the same thing on Prem or different clouds um you know you can use EFA that that particular example doesn't use EFA but if you need a high-speed interconnect uh you know you can use that as well so yeah it's pretty pretty flexible and I you know once

again I get excited about when I see our work uh having real world impact and helping people and so I'm I'm excited to see this this do the same and keep on you know carrying that through the open source community that was awesome Dave I also just want to give another little shout out to um alter is going to actually be in our booth at sc23 on Wednesday the whatever that is I think it's the 15th I think so November yeah we don't have a was that I said that sounds right now you're gonna make me look yeah yeah I think that's right I think

that's right it is you're right um we don't have an exact time yet but you guys are actually going to be in our ciq booth so 18 55 doing a a demonstration with fuzzball so I think that's going to be really exciting to see somebody not ciq getting in there and playing around with fuzzball and um you know hopefully it goes smoothly and then also there's a little part of me that's like break everything so that we can fix it and make it better you know and like that's kind of kind of the the the fun of it so very exciting we're very excited to

see you guys and um Eric I do actually have a question for you so hopefully I'm gonna pronounce all of this right and we can go with that because sometimes a little Tech techy type stuff I'm like oh go off on a tangent so I've heard that ciq um used some test model in LS Dina um input format can you tell us a little bit more about that yes that that's a good question because this is what we we saw today in fact because for let's say specialist you recognize the star which is let's say LD format so the the idea is that let's say

since a few versions radios and now open radios is able to deal natively with the lsda format so we understand that this other solver is very popular okay okay so many Engineers researchers already know this format so we want it to to lower let's say the entry barrier to use open radios and to to support natively this popular format participate to make open radios I would say more accessible to a wider community so not have those who knows radios format but also those who knows this other format and they they can just take their models and run it with a with open radios okay without

modification of the the input deck so I think that's uh that's really great to be able to start and to continue to use let's say the the way you you know how to model models but using another solver and in fact we were able to do that because uh indeed LS dinina on radios Shar the same rout which was the Dina 3D research code in the 80s this was at that time that the one of the first let's say explicit code was was built and then there were several let's say different codes spread over the world they evolve let's say in parallel but they had

they had quite the similar functionalities so it was more a way to to be able to understand this language and to interpret it with with radios natively and this is what we have been able to do and now we are helped also by the community to continue to spread and to support more more more features with this let's say alternative format that's awesome yeah thank you I know was this point earlier we're actually GNA try to stream all of those demos that we do in the booth too so if you can't be there and you can't see out there come back to the channel we're

gonna try I've heard I'm not gonna put too much stress on everybody and say we are going to because I know it's a lot of effort and I know we're doing our best to try but we're gonna try we are gonna try that's awesome so I know we're we're pretty much coming up on on on time here and I really want to uh respect your time and Eric I'm so so grateful for you being here I don't even know what time of the day it is for you but I imagine that it's not 11 in the afternoon like it is for me yes it's a

little bit later but that's that's good you know generally during work hour I am really occupied and busy so in the evening like that it's it's easier for me to to be available and to participate to to great a a webinar like this one it was really my pleasure again I want to to thank ciq for this opportunity and for his support of of our project of this community and all the work all the effort to democratize and participate to spread open radios by making it even more accessible so big thank you to you and my pleasure to be there oh and thank you thank

you to you Eric and alter as well yeah it's uh this has been a lot of fun working on this and I I look forward to continuing to see us work together on this yeah absolutely thank you very much awesome all right you guys so thank you everybody for watching thank you for sharing this amazing demo we got two little Demos in there so that was really cool we got to actually see it in action then see fuzzball in action we've got the backstory to alter and open radios and it was just absolutely an awesome a great show so thanks for listening on our podcast

thanks for watching on our YouTube um and make sure that you come back so same time same place we will be here next week thank you very much again Eric for your time and uh Dave and z you're amazing have a great week bye thank you Rose Dave Zan and see you in couple of weeks in SC in Denver that's it see you there see you good morning good afternoon and good evening Wherever You [Music] Are