Astra and Sandia's 725E Data Center

[Music] Vanguard is a program to investigate advanced architectures and advanced technologies and their viability in supporting the NNSA mission codes specifically Astra is a project the first project within the Vanguard program Astra was selected from the phrase per aspera ad astra which means to the stars with difficulties so Astra basically means star astra is going to investigate the viability of using 64-bit high performance computing ARM processors for the NNSA mission previously ARM processors have been around for a very long time they're very ubiquitous so we have them in our cars we have them in our cell phones but up until recently we haven't had arm processors or arm-based processors that were viable for use in our HPC mission astral will be the largest arm based processor cluster and we'll be using that to prove out the viability the NNSA mission codes so Astro will have a theoretical peak performance of 2.3 petaflop so while that's that's very fast what is most important is the percentage of that theoretical peak that we can achieve for our actual mission applications that will tell us whether or not Vanguard is viable far and in the same mission applications Sandia was chosen to host the Vanguard program in the astral project in particular because of our long history with advanced architectures we've had an advanced architecture testbed program now since 2010 where we investigate a very broad selection of technologies and their applicability towards high-performance computing in general in the NNSA mission the advanced architecture testbed program has allowed us to form very close relationships with OEMs and technology providers in general so this has allowed us to have a good understanding and a good influence over future technologies arm has emerged into this this pool of advanced technologies that we can hopefully leverage for our NNSA mission one of the things that we weren't able to do with the testbed program is actually test things out large-scale these were typically investigations that were somewhere between 16 and 32 nodes to really understand the viability of a technology for the NNSA mission it's important to do large scale Akilah sandy is the best choice for astra largely because of its our legacy of research and we have some of the best people in the world best capabilities in the world our facilities folks have had their eyes towards the next-generation technologies for data centers and the planning of this facility has enabled us to serve Astra well for building the very first big arm system it's great for sandy as a whole and to host it in this LEED Gold Data Center is just a plus so when you talk about LEED Gold buildings it's a point system so basically you have to be very energy efficient and you have to have a lot of good infrastructure built into this sustaining infrastructure this is sandy is very first LEED for building and it happened to be our data center the design of the data center should be a data center property so it's not fancy it's very functional every piece of floor space is usable there's no spans in the middle so we're able to have a continuation of racks and very limited by candy in the data center is basically a functional data center we're gonna save millions of gallons per year of water just by this thermo siphon which is a very efficient way to do a heat exchange of the hot water that comes out on a big HPC data center it's pretty cool technology at basically works off phase change of refrigerator there's no pumps it just says fans you take the warm water and to the thermal siphon and refrigerant changes stage rises up and it's cooled off by some fans and it drops back down and cools the water entering it's pretty neat we expect to save somewhere in neighborhood of 6 to 18 million gallons of water a year once we're fully deployed so Astra is water-cooled indirect system and it fits perfectly and the design for this data center has oils internally fans internally and our closed-loop system works well with it so we'll basically only have about 8% air cooling and the rest like we're cooling on the system on the air side hot air rises so we allow the hot air to rise and just say collectively up there and we take out the hot air that we don't need the rest of hot air just stays as a cushion up there insulator if you will and we leave that up there it circulates back down it pushes that hot air up there so gives us a more open look and a better aerodynamic situation when 725 was initially constructed for Red Storm it was constructed with expansion in mind being able to just add on to that building was very convenient for us and we can really expand our ability to host high performance computing platforms beyond Astra or maybe in Vanguard phase 2 whatever we choose for Vanguard phase 2 will will likely be put in 725 East [Music] having the system delivered while the building was still under construction that's that was a significant than ever it was a horserace basically so in their interim were developing the astra system and getting bidders to bid on the system and at the same time we're designing and building the data center so we knew it was on a collision course so all our folks were safety trained and wear hardhats an appropriate protective gear and always be aware of construction activities around us sort of like you're building a plane or cargo plane while it's flying fortunately we were able to deliver the machine well the final construction activity we're still going on so everyone has a role like the Eunice's team their cabling and put in the machine components together all the teams are well in the racks in place setting them in place then my team takes over and connects all the infrastructure the cooling the power making sure the floors intact all the cutouts are done my team's gonna operate the machine over the life of the project from the start we were part of the RFP F for architects and folks within my team that were either on the IO or the system software side we're part of evaluation and a selection for the Astro machine and then as it came closer we had several folks within my group working on the system software stack which is gonna be the operating system that drives the machine and we had our storage folks io folks working close make sure that they'll have sufficient IO to deliver the cycles or deliver the mission of Astra then we also had our Hardware folks getting involved so they're familiar with the hardware infrastructure and how the service and help maintain the platform to have our hoaxes that manage the systems day to day to fully understand the machine and be able to help users and help diagnose for help point why not hardware problems or help keep the Machine functional I'm real excited about Astra and its role within Sandia within the nation and within the world in the sense of it's really opening up a whole new platform I feel that this is a milestone for Sandia this is the cream on the coffee here [Music] [Music]

Channel: Sandia National Labs

Views: 20,815

Keywords: Sandia National Laboratories, supercomputing, supercomputer, hpc, astra, data center, leed gold, leed, arm processor

Id: 6tneOq5M9m4

Length: 7min 51sec (471 seconds)

Published: Fri Apr 24 2020