 Yeah, we can ask you from the people Hello test Hello, can you say something? Hello Okay, is our audio balanced? Okay, so I'm still getting set up here at least Do we have anyone watching us? I think we have few people What kind of icebreaker should we have? for today Let's see Maybe something yeah Well, you'd ask things like what kind of parallelism you use or Like given what you know now, how do you use the cluster or bigger reason? Rico had a thought Do you think your work could be parallelized? That sounds good. Yeah. Yeah Yeah, today. We're going to have a lot of talk about the Pack a lot parallelizing your work so so the like the HBC like the high-performance probably of the high-performance computing It's interesting to see what what people have Do remember to give us feedback also at the end of the day like the feedback is the only way that we can like Decipher whether whether what we're doing is is good or not Also, we would like to plug many other resources at the start here today. So Yesterday, we already did a lot of work with With a command line for example Which is a tool that is Invaluable like it was mentioned in the first day quite a bit the command line It's it's very it's invaluable when it comes to like scientific computing and especially a high-performance Computing because that's like the direct way of talking with the machine Without any like any Intermediaries so Learning about it is very Very good idea Especially if you know that you're going to be doing this for a longer longer time. So There's plenty of good resources on that. So our own bash course is really good like You can you can take it and it's it's Really really fast really good way of getting yourself accustomed to the command line a set up Then there's also like great resources on the on CSC side on on cluster usage and then also Code refinery has great resources on it Uses project management all kinds of stuff like that and also the upcoming code refiner course coming in March, right? Yes, it's actually on the code refinery web page Yeah, so we might want to link on that in the so so there will be well everything you might Want to know about like how to how to get better at at coding and managing your workflow and managing like learning about notebooks learning about version control How to how to do stuff efficiently? It's very good like because then you might These like doing these hands-on things it might be easier to to get actually a Custom to the to these tools that we are using Yeah Did you mention the research software engineer service yet? No, you can do mention. Yeah, so Like we sort of implied on the first day We are covering a lot of different things and there's basically like I mean there's so much that a person might need to know And we realized that not everyone is able to like not everyone needs to or Has time to learn everything in the amount of time they have So That's the idea behind this service here, which Here so In the service We basically have people like us here who go and directly help your work So instead of saying, okay, here's what you should do. Good luck. Go figure it out We can go and directly improve your things so we can do the software programming for you if you need to we can Automate the workflow we can help you scale it to try to and we can optimize it and so on And then this is really important because I mean there's so many different ways of doing computing and so many different styles like it's not just the People that have been messing with Linux and That kind of stuff for many years that need to do this So we allow everyone to be able to get stuff done. I guess a good way to Get started is our daily garage So if you're at Alto every day at 1 o'clock, we're in a zoom meeting Well, I mean not today and during this course actually there have been people there during this course But we're basically there you can drop by and ask us any simple question. So you can say I'm doing such and such is this a good idea Should I ask for your help to do more? And so on Also, I'm pretty certain that other universities have similar things going on these are getting more popular because these kinds of Incent like organizations because the There's so much of this coding and best practice and all kinds of stuff that is related to scientific reasons But it's not related to the science necessarily. So it's good to ask people who are specialized in these kinds of like coding things and best practices and and like to ask ask help for any kinds of like problems that you might have before you start a protocol when you're Stocking your project So, but yeah, we're We're it's time. Yeah, I think we are ready to start Okay Yeah, yeah So should we quickly summarize what we did yesterday, right? Yeah Let's see Yeah, so yesterday where did we start So yesterday we went to the actual cluster. We run some commands interactively so that they would run on the compute nodes instead of just the login node But then the meat of the day was that we run these serial jobs and the serial jobs were Like like scripts that we wrote the instructions for the compute nodes run when they went through the queue So we had the queue system that Manages the job allocations and the serial jobs were then the instructions of what the jobs were Running when they well ended up running eventually on on a compute and we tested out different kinds of serial jobs And now we are basically like moving Using that basic framework of idea of like we have this script that that we give to the queue system and then we push it to the Queue and run it on the compute node. We make that into Well more advanced stuff today So, yeah, I guess we have three ways of parallelizing things overall Yeah, so so there's a few ways and we have a first the array jobs, which is this kind of like What's it called I forgot even like Embarrassingly parallel in that embarrassing to forget that but for the embarrassingly parallel way of parallelizing things or like basically like adding more More people doing the basically adding more pots to the to the pasta analogy that we had and then we have these actual parallel computations That are basically like we want to see individual jobs. We want to give it more resources to run faster. So we give it more GPUs to run and then we have the GPU Resources that are like they are parallel Calculations done by the GPU, which is this processing unit that can do this parallel calculations in the in the Like separate card accelerator And in the middle, we have a nice talk from CSE Where we will be talking about how well they will be like introducing the Services and telling how you can scale up to even bigger systems and how you should how you can get even more resources in the national level Yeah, I mean so you might think that Triton is large but at the scale of CSE it's not very large But also then there is more like Effort moving things from place to place to do that. So, yeah Yeah Should we stop with the right jobs? Yeah Okay