I Don't Want to Live On This Planet Anymore: Epyc, Find Us a New OneFeb 22, 2022
anymoreand you know mars mars isn't exactly attractive looking it's really possible there aren't any other easily habitable globes in the skies surely there is one somewhere for every grain of sand on earth's beaches there is at least 10 stars in the universe and you think there is at least one
planeton average around those stars, if not more, so there will be more planets than stars, surely there has to be something somewhere, an extra solar earth, so what? Where are everybody? those planets we need to
findthem scientists are developing new techniques to
findthese extra solar planets all the time i mean we have a lot of data we just don't know how to reliably find extra solar lands in the data we have it's a problem complicated.
One technique involved the development of a coronagraph that revealed four Jupiter-sized planets orbiting a star. Now
thistook almost a decade of observations. We don't have that kind of time. another method is called the transit method basically what it does is it looks for a star and then it looks for the star to dim periodically and based on how much it dims and how often it dims it can figure out how big the planet is you you know where it is and the really cool thing about this technique is that you can use image arrays and image a bunch of stars at once just take a bunch of observations you can make thousands or tens of thousands of stars whatever you can fit into a field of view so this lends itself to uh numerical analysis data analysis big data analysis or whatever you
wantto call it to find additional solar lands however the problem is the land isn't really that big so the amount of dimming you will see from planet size to earth yes you know stars are analog creatures that will dim and brighten on their own enter lowel observatory l the lowell observatory is an independent non-profit research institution founded in 1894 by bost about the mathematician percival lowell is near flagstaff arizona scientists from the university of lowell and yale university have built something they call the precision spectrograph This is one of the first instruments where scientists really hoped to have the level of precision needed to locate extrasolar lands that meet the criteria and that, uh, you know, they're not gas super giants or anything like that, but they do have an effect of local dimming quite dramatic now at Lowell observatory they're already collecting lots and lots of data, in fact they're having a bit of a data overflow problem it turns out if you can't process a day's worth of data in less than a day, usually much less than a day, very soon you will be overwhelmed with data, drown in data the data you really need It needs to be processed quickly and also distributed to other yale university research scientists in the world, they are not just sitting around, they need to process the data in o To do their analysis and research, the data overcomes a huge mess of data drowning in data .
Sounds like they could use a computer janitor so I've been talking to Dr. Joe Lama about what kind of computer janitor services they need precisely because I'm on their kill so while astronomers are looking for planets around other stars at night, we still need a good model of how stars behave. a better model of our sun could help us do that, so this instrumentation is used during the day and night to collect different types of data, dr. lama is running the experiment for the solar collection aspect and nasa has already collected 2.8 petabytes of data so the hope is that this massive volume of data is just the first chapter in an epic story about human space exploration and the search for the first habitable exoplanet plan but I have to call in some favors yeah yeah yeah yeah yeah well I mean you know it's for a good cause they're a nonprofit we really could I mean you know we really could do a lot oh great thank you that's that awesome I really appreciate it I'm sure they will too thank you thank you thank you again good news everyone well I have a plan and my plan is to process that data as close to real time as possible as soon as possible. the data goes into the process and gets an indexed compressed version of the data stored on a high speed storage array something like a flash drive now we are going to keep the original data on a hard drive array that is local to the server so we could quickly load data from that mechanical storage because hey 18 terabyte hard drives are a thing and 18 terabyte flash drives are expensive so we can reload any data we need from the local storage array if we need to order from even more storage can fit in a single chassis because 200 gigabytes a day, don't you know?
Even 18 terabytes don't last that long or 18 terabytes multiplied by 12 drives, we can offload that additional storage for networks and also store the information. which we process as close to real time as possible, we'll probably also implement a couple of containers on the server so that other research scientists can make a request and fill out a web form basically and then be able to get whatever other information they need and download it in real time . This will likely be an ongoing process. You'll probably end up watching a couple of videos on this and the other stuff that goes into it, so now let's wait, wow, that was stealthy.
I think it's time for a build. do what your mount voltron entailed gigabyte and amd have really contributed generously here we can't thank them enough without them this project would not have been possible and kiosia also went
livefor some flash storage we are using eir cm6 Enterprise ssds for our storage array and we have four of them set up in this chassis so we will be able to process data as close to real time as possible thanks to this flash array the final build has two 32 core epic processors and a terabyte of ram some of you at home have SSDs out of 512 gigabytes this computer has a terabyte of ram now with this hardware we can process a full day's worth of data in much less than a day so we can catch up with all the data that trumps everything else and can also reprocess data very quickly every time they change the algorithm, the centerpiece and crown jewel of the plan here is of course the gigabyte g482 this is a research scientist's dream server it's for you you know most of the time the data center is looking for the most compact machine possible but this chassis is one of the most flexible enterprise chassis I've ever viewed. a storage and we also have all the pci express connectivity that we would need and we don't have an algorithm yet that takes advantage of gpus or fpgas or anything like that so all the heavy lifting here is down to amd epic and the monster 32 core cpus that we have in this chassis are going to slice through that data real fast no problem 32 core cpu balances high clock speed and core density for sort of a sweet spot for this particular algorithm looking for extra sun lands now still plenty room for optimizations in fine tuning in future videos with level one, but let's check in with Dr.
Lama and find out a little more about Lowell Observatory, his team, and what's going on so we can have fun and live vicariously, so I have al Dr. Joe Lama here with me. Hello and welcome. You are live from the observatory. Hello Wendell. amount of horsepower and you're collecting data almost faster than we can actually store it because you know you and I were working together and looking at some stuff and that kind of thing, but uh, you know, tell us how the deal goes with this level of data it's absolutely bananas and you're using the telescope during the day yeah I'm doing everything backwards I'm definitely a social hour astronomer not a night astronomer so yeah you know when we started this project We don't really think too much about you.
I know the computing needs of this, it was kind of like, oh, here's a cool idea, let's do it and we'll figure out the details later and into, you know, as you well know, those things come back and shape you, uh, so, but yeah . you know we have this star during the day the sun uh no none of my colleagues are competing for telescope time they will just want to use night time so i have unobstructed access to the telescope all day every day more or less uh then yeah and as you say that generates there's a lot of data well i mean it wasn't just you though i mean it was like i was doing some of the prep material for this and nasa and one of the things you sent me is like oh yeah nasa just casually dropped in oh yeah we have like a petabyte of data on the sun during daily observation and okay there's something to this oh yeah absolutely so the sun is just you know it's so as close as you know all the other stars you see in the night sky are just these point sources and while we can collect a lot of data about them they are still just point sources and a lot happens between us and that s stars that prevent us from obtaining as much data as possible about them. so you know stars like the sun is this really cool opportunity to look at and just understand what's going on inside these big firefighters is really a cool opportunity yeah i think you know we've already been testing the server a little bit with some data sets and you know it's going through it and um you know it really is like a buzz saw just cutting through the data i mean amd epic and what gigabyte has put together uh on this chassis they really drive it home as far as the data processing. in real time and like whatever crazy software and algorithms and stuff they come up with, it's really crazy to think that this could be the machine that finds out that you know a planet where and it's like, oh, we need to take a closer look at this, oh we're getting polarized light from this planet maybe the plants on the planet are absorbing the other polarization of light you know it's possible oh my gosh yes it's so true um yes we can't thank you enough to a gigabyte, this is much more computational. power we never thought we'd have access to or maybe even need, but you know, when someone gives you a hammer, you find the nail to solve the problem, so we'll throw everything we've got.
I have on this machine now I'm really excited you know it's not just that you're not just collecting data for yourselves not just in the project either I mean you have a lot of other higher education researchers and other researchers that you're feeding the data to data so it's not just you Know your pool this depends on I mean there were some pretty big names in there yeah that's right so what we want to do is you know to make this data as open as possible so all my colleagues around the world can know download it and try to overcome this problem that we have of trying to tease out the little signals from the orbiting planets, yeah you said um I think in the existing system, when you're running the processing of data and you like to open the terminal you would type some stuff and then like a few seconds later it would go through the screen it's like using dial up internet that's the process or thought i had oh load average is 150 that's good yeah so yeah this does quite a bit this this will traverse t hat with uh uh a little bit easier than uh than the existing setup and we can scale it as well in terms of making the data accessible and dealing with that on the software side um we're looking at doing that with containers and that's something we can go over in a follow up video.
I'm sure I'll try to make it public at some point so there will be a part 2 at some point, yeah definitely comment below and you know. the whole thing and we'll see if we can, if we can, you know fix that, but I'd love to go and see things in person and, like, let's shred the data, do the consulting thing and find out, you know? and it's really cool to have such a powerful machine that you know what it does to you normally when you want to reanalyze something you know you change something in the code and then you have to wait a week to see if what you did works or not and with this system it doesn't we're going to have to do that so you're going to know j aside it will really speed up literally both the data processing and the parsing it's so cool yeah both sockets are i think you'll be able to do your processing in real time like normal but also when you have an algorithm change that can run in the background because you know you have months worth of data and that will still take a while to process but it won't interfere with daily processing and this chassis because it has a lot of pci express connectivityyou can also expand the physical disk storage so you can have cheap local attached storage you know we did our garbage server build and you could use something like that to add another 48 base three and a half inches and that will take you up the range of petabytes with some reasonably sized hard drives and that's local attached storage so it's automatically very fast you can do stuff over the network of course it's also more interconnects but also fpga s and gpus at some point in the future it's like you have to invent the algorithm on cpus you have to do the invention and hard thinking with general purpose computing you can't do it until you figure out what you need with something specialized like a gpu and that's why these things are so critical to research scientists, yes, that's right, and they know that people, my colleagues are starting to develop all these machine learning techniques omatic and all these cool applications that we can use gpus for so this chassis is going to be I'm going to do it all yeah it's really going to be my one stop shop I don't have to buy these really powerful desktops
anymoreyou know what i can get like my fancy imac or whatever and zoom it in it's going to be great yeah i'll be able to run all your jobs too i mean you know you might have an award from the office for the person who submits the job that actually crashes the server , yes, that's not bad, we could do that competition, I think. guys have a joke ed i think i have the most powerful machine probably in northern arizona maybe so yeah it will be great well epic will definitely bring home the bacon no question about it so i will . interesting to follow up, well thanks for coming, it's been amazing, any wise parting words, anything you want to see.
Yeah, lots of awesome stuff. We are a nonprofit research organization. You know we are doing the best we can in these difficult times. Book a tour thank you so much to you and the gigabyte guys for making this happen it's beyond amazing I'm so excited how it all worked out oh amd and keoksia I called in a lot of favors. I'm going to have to land right negotiating with you it's pretty easy when you're a non-profit because you can use the magic words of tax deduction and then you know all of a sudden all the doors open and it's kind of magical but yeah, you know observatory lol look at it I really love the work that's being done and it's really exciting to think about the next one you know uh uh what I always say is an extrasolar earth but it's uh what do they call it?
We just shortened it to exoplanets, oh yeah, that's a, you know. Earth-like exoplanet is kind of how we say it in the business. The next Earth-like exoplanet could be found by AMD Epic and this machine that's pretty crazy, so it's really awesome to get the chance to work on this, so thanks for that. oh well thank you this has been a great project it's been a lot of fun and this is also just the beginning so I'm looking forward to more of these great ones well once again I want to thank amd and gigabyte for their amazing support on this project i love working with nonprofits if you have something you want to do you know reach out maybe we can work something out it's been really great special thanks to keoxia for their flash storage array i'm sure this one isn't its the first time we are going to see this gigabyte monster beast chassis and server videos because its pretty much the chassis to have in a hyperconverged infrastructure i mean it does everything be sure to check out lowell observatory check its mission and maybe I paid them a few bucks, after all, they are a non-profit organization.
I'm small. This is level one. I am disconnecting. You can find me on the level one forums.
If you have any copyright issue, please Contact