Meet El Capitan, “The Most Powerful Computing Machine in the World”
How do you build what aims to be “the most powerful computing machine in the world”? For starters, you can watch this video to see how Lawrence Livermore National Laboratory is constructing its next-gen supercomputer El Capitan, which will be deployed in 2024. Through the lifespan of the machine, lab employees will need to operate, troubleshoot, and maintain El Capitan around the clock, seven days a week, ensuring that scientists, physicists, and code teams can perform their calculations efficiently and in a timely manner.
Transcript
00:00:00 elcapitan is the national nuclear security administration's first exoscale supercomputer lawence Livermore National Laboratory has been deploying worldclass supercomputers um since the 1950s and every time we get a new generation of supercomputer it allows us to process more information faster with greater detail uh to accomplish our scientific Mission the neus scale supercomputer
00:00:25 doesn't just stand by itself it has to sit in a facility that is capable of of operating it the Cornerstone elements of supercomputing in facilities are that you have to have first the space the square footage you have to have power you have to have Cooling and you also have to have um the structural Integrity of the facility to support the infrastructure building 453 is gone
00:00:46 through a lot of different electrical changes um originally when the building was built is for terascale computing is for 208 power um The Next Step went to um pedis scale Computing and that's a requirement of more 480 volt power so we've transitioned from more uh commercial uh solutions to Industrial solutions to now utility Solutions so the latest um infrastructure upgrade is
00:01:10 called exoscale Computing facility monitorization and that is actually bringing in a total of 85 megawatt of power into building 453 as well as 20,000 tons of cooling underneath the floor we have electrical and cooling so there's big pipes that bring our facility water to the cooling dist distribution units and there's a bunch of electrical infrastructure that
00:01:34 eventually lets us plug the big line cords um into the the racks for power the networking on this system actually runs over the top the amount of networking cables used in order to wire up Ayan if we were to lay it out would be multiple football fields long the sheer size of networking involved for a system like elcy 10an involves several people working around the clock just to
00:01:56 lay down the cables for the networking so just in Livermore Computing we have have about 120 people uh who who work on on all of our systems but most of them have been involved in this one because it's such a significant effort for us and then on top of that you have hundreds of additional people between uh the vendor for the system as well as contractors for electrical and
00:02:17 mechanical um other folks at the lab who aren't directly liore Computing but support us so it's really just a huge uh you know group effort to make this thing go in operations currently there is a sense of excitement because we haven't been ble to touch the the machines yet um I think everyone's excited to to open her up and see what the engine looks like right and and be able to touch and
00:02:39 feel everything and make sure uh we understand um how the components work and on a hardware side how we can be effective and um use our skills to to troubleshoot any issues moving forward so we're working directly with the vendor this time and out the gate we're learning side by side with them all the different components and all the tools that they're they're learning how to use
00:03:02 we're learning right alongside with them so there's a lot of uh unique things about elcap that we're learning um and as with every HBC high performance Computing system we have to learn how to monitor it uh not everyone is monitored the same we don't know what we don't know yet as far as how we're going to tie their monitoring tools into our monitoring tools um all of this we're
00:03:22 going to learn together in preparation for ELC Capitan we received Early Access machines EAS systems these are predecessors to the elcapitan architecture that our application developers use to ready their applications they're our best look at the architecture that elcapitan will be in the absence of any of that system being available to us though these E
00:03:43 Systems are a fraction of a Capitan they're already rival the top 200 world's fastest supercomputers I feel really excited about the process and all the challenges coming in order to get this exco computer operational and to see what it will bring to the world in the realm of science I love working on this stuff um I have been at the lab for about 20 years now and almost all that
00:04:05 time I've worked in the hiformance Computing Center and it's just so cool to work on the fastest computers in the world if you think about every person on the Earth all 8 billion of us if every person on the earth did a calculation an addition a subtraction every second of every day of every week of every month of every year it would take the entire Earth 8 years to do what El Capitan will
00:04:27 be able to do in one second I cannot wait to see what El Capitan allows us to do the reason I work at a place like Livermore that Fields machines like this is because you cannot do this anywhere else elcapitan and its sister machine twam will allow our scientists like myself to do things they could only dream about 5 10 years ago that's the reason I come to
00:04:51 work