National Energy Research Scientific Computing Center (NERSC) Perlmutter //Science Accelerated, 27 May 2021

Previous Meeting

Next Meeting

⏯

youtube image

►

From YouTube: Perlmutter Supercomputer Dedication

Description

On May 27, 2021, Berkeley Lab and the NERSC dedicated their newest flagship supercomputer. The HPE Cray system, dubbed "Perlmutter" is built to support a wide array of science workloads, from simulation to data analysis and artificial intelligence methods.

A

Thanks for being here today., this is a very important moment for the berkeley lab community.

A

And we're so pleased that all of you could join us today. .

A

I want to take a moment to welcome distinguished guests from the department of energy,.

A

Our industry partners, nvidia,, hpe, and amd, and our university of california partners.

A

We are here today because the department of energy's strategic, long-term investments in.

A

High performance computing., this farsightedness has helped our nation to be prepared for both.

A

Exascale and quantum computing and to expand the role of hpc high-performance computing.

A

At berkeley lkab, our strategic plan for enabling scientific discovery recognizes that.

A

Computing has transformed nearly every aspect of scientific inquiry across disciplines and across.

A

Scales from the behavior of subatomic particles to the formation of structures in the early universe,.

A

From the assembly of the human genome to the evolution of earth systems, this deep integration.

A

Of computing into all aspects of our research make it possible for us to sustain a leading role.

A

In the transformation of science that is resulting from deeper applications of computing.

A

And mathematics so today we're here to celebrate the installation of our new hbc system, which will.

A

Provide a dramatic boost in computing performance and capability, not only for the labs researchers.

A

But for scientists from across the department of energy in a nod to the extraordinary research that.

A

Has come from the integration of computing and other scientific disciplines? We have.

A

Named this new system, mattress all perimeter, my colleague for his nobel prize winning work saul's.

A

Partnership with nurse has been a testament to the effective and productive collaboration.

A

Between cosmology and computing science, and is a compelling example of how researchers across.

A

Many fields can harness new computational approaches to accelerate the pace of science.

A

Now, I'd like to introduce our keynote speaker, the newly appointed deputy secretary of the.

A

Department of energy david turk, david, has spent his career in public service, has held prestigious.

A

Positions in both houses of congress, the department of state and the white house.

A

And this is his second tour of duty at the doe. He previously was deputy assistant secretary in the.

A

Obama administration, up until his current appointment david was the head of the energy.

A

Environment. Division for the international energy agency and we're pleased to be working with him.

A

Again, he has deep expertise in clean energy and a passion for the transition to a low-carbon economy.

A

I know I speak for our entire lab community. When I say we stand ready to help you in this effort, as.

A

You begin your tenure. Please join me in welcoming deputy secretary turk. Thank you. uh Thank you very.

B

Much mike for that very kind introduction, and, more importantly, thank you for your outstanding.

B

Leadership uh and for inviting me here to join you all uh for this virtual incredibly.

B

uh Special event and good morning to uh to all of you, today's dedication of the new supercomputer.

B

Is a historic milestone which is just adding to the nearly 90 years of scientific accomplishments.

B

At the berkeley lab incredible accomplishments like demonstrating the health risks of radon gas.

B

Leading to epa standards, creating crispr genetic engineering tool, advancing our understanding of.

B

The causes of breast cancer, explaining how photosynthesis works and capturing the first.

B

Microscopy images of the smallest bacteria and I could go on and on that's quite a record of.

B

Groundbreaking groundbreaking accomplishments and congratulations to all of our participants who.

B

Were involved in any of those that I mentioned or any of the countless others that I did not.

B

Mention and of course lbnl is home to 14 nobel laureates, and I know from the big bang theory.

B

Which is my 12 year old, son's favorite tv show that one of those nobel prizes belongs to dr salt.

B

Pearl mudder, who we're celebrating today, saul, I'm wondering if you take more pride in discovering.

B

The accelerating expansion of the universe or in triggering such professional jealousy in dr.

B

Sheldon cooper we'll wait for your remarks for the answer to that very important question.

B

Visionary scientist like dr paul mutter brings me and I think, all of us immense pride not only.

B

In berkeley lab not only in the department of energy and in our country, but in all humanity,.

B

And please rest assured, secretary granholm myself, all of us in this administration and the biden.

B

Administration feel to our bones that america, our entire world, are much much better off we're safer.

B

We're more prosperous, we're more inspired when we invest in science and when we invest in innovation.

B

And to solve today's most important challenges, we need to invest in science and innovation.

B

Like never before, there is simply no way our world is going to be successful at achieving.

B

For example, net zero emissions in just a few decades without a continued stream.

B

Of cutting-edge lbnl solutions, building on the incredibly uh impactful work that has already.

B

Been done on cool roofs, next generation, ev batteries, for example, and our world does.

B

Face a daunting array of challenges, which is why we so critically need the pearl mudder.

B

Super computer: we need the supercomputer to help us track extreme weather events to help.

B

Us respond and adapt to intensifying climate impacts. Its number crunching ability can help.

B

Us explore our solar system find new treatments for diseases. It can help us discover new materials.

B

Develop new energy sources and advance research on artificial intelligence. But- and I want to.

B

Underscore this certainly especially for some of our younger colleagues who are joining us today,.

B

No matter how cutting edge our super computers and our other facilities and technologies are what's.

B

Most important are the people behind this cutting edge science. Let me think again, not only saul.

B

uh But all the scientists who are here today for your public service, for your passion and for your.

B

Teamwork and we'll certainly do everything we can here in washington dc to support all that.

B

You're doing at berkeley lab and those who are working at our labs and universities across the.

B

Country, the president's fiscal year 2022 budget, which the secretary of energy is testifying right.

B

Now in front of the house, science committee and the full budget will be released tomorrow,.

B

Calls for over 8 billion us dollars in clean energy technologies and innovation, 700 million.

B

For arpa-e and the new creation of a new rpc 7.4 billion in the office of science in our national.

B

Labs president biden's america, jobs plan calls for an additional dedicated reserve of 20 billion.

B

Dollars for upgrading research infrastructure at lbnl and our other 16 national labs across the.

B

Country and we will put even more emphasis day in and day out on, attracting a talented and diverse.

B

Group of scientists and other professionals dedicated to public service earlier this month.

B

For example, secretary granholm announced 17.3 million in funding from the department.

B

For college internships and research to invest in historically black colleges and universities,.

B

And other minority serving institutions, we need to make sure that young people of color people from.

B

Low-Income backgrounds, all kinds of backgrounds can get in the door and succeed in places like.

B

Berkeley lab with its 90 years, almost 90 years of incredible scientific accomplishments. So.

B

Congratulations again mike to you to all our lb l colleagues and to everyone who's involved.

B

In this historic achievement, and again, thank you for having me here with you all uh today.

B

I'll turn it now back to you mike, so thank you david uh for those inspiring remarks. Our next.

A

Speaker, steve binkley needs no introduction to many people here. Some people knew him back when.

A

He worked at sandia, national laboratory and many more from his time as head of oscar. The.

A

Advanced scientific competing research program at the does office of science.

A

And in that role he was instrumental in mapping out the strategic roadmap for the.

A

Growth and expansion of the does high performance computing program that we're benefiting from today.

A

So now, in his role as principal deputy director for the office of science, steve brings his.

A

Considerable experience and deep knowledge of the scientific resources at due and the national labs.

A

To ensure the agency has the resources it needs to meet its mission goals.

A

His strategic leadership in this role and others has been especially important.

A

To the success of cross-cutting programs that affect more than one program office,.

A

Along with special research initiatives of priority to the director in the department.

A

Leadership, please join me now in welcoming dr steve binkley to the virtual stage.

C

Okay, thank you mike for the very kind introduction.

C

And it's really a pleasure and an honor to be here participating in this dedication today.

C

It's taken a long time to get to this point.

C

The success of a nurse at berkeley is really something to have seen over the last 20 odd years.

C

It also is a testament to the skill and the professionalism of the many employees at.

C

The lab that have been involved in this and to mention just a few horse, simon and kathy ellick.

C

Who played a key role, getting us to exascale and also it's very, very fitting that the.

C

New computer is being named after saul, pearl nutter for his pioneering advancements in.

C

In cosmology, and also his involvement in computational science uh in place.

C

At nursk, so with that I'll conclude and again, thank you for the opportunity to be here today.

C

Over to you, thank you, dr binkley, for those kind words, um hello, everyone, I'm jonathan carter,.

D

Berkeley labs associate director for computing sciences, um welcome everyone and thank you for.

D

Helping us celebrate this milestone for nurse for berkeley, lab and department of energy office of.

D

Science, here at berkeley lab, we believe some of the biggest contributions to discovery.

D

Occur at the intersection of diverse disciplines, ideas, expertise and instruments for nearly 30.

D

Years, the computing sciences area has been the nexus for many of these collaborations in the.

D

Lab and across the doe contributing our expertise in computational science, mathematics, networking.

D

And high performance computing to help scientists from a variety of fields solve some of society's.

D

Greatest challenges, as we look to the future of science, increasingly sophisticated instruments.

D

From supercomputers to telescopes to light sources will provide unprecedented insights.

D

Into our natural world, the expertise of massed in better labs, computing sciences, area.

D

Combined with powerful next-generation supercomputers like the pelmeter system,.

D

Will allow scientists to run increasingly more accurate simulations and apply cutting edge.

D

Analysis techniques: to achieve the critical breakthroughs we need to address.

D

The many challenges that we face as we envision this future I'd like to take take this opportunity.

D

To recognize one of our biggest champions over the years associate director of the office of sciences.

D

Advanced scientific computing research program, barbara helland, we've been working with bob for.

D

Over 20 years, first, in her role as ask as oscar hpc program manager, then as facilities director.

D

And now, as associate director, she's consistently and strongly advocated,.

D

For supercomputing facilities, enabling them to deliver world-class computational resources.

D

For the does mission, and also to sponsor the research and develop necessary to make effective.

D

Use of these systems, so please well join me in welcoming barbara hello. Thank you. Jonathan.

E

And I want to add my congratulations to berkeley lab and nurse for today's launching of perlmutter.

E

I'm looking forward to adding to the adding it to oscar's arsenal of high performance computers.

E

As many of you are aware, pulmuter features a heterogeneous architecture that will provide four.

E

Times the computational power currently available at nursk, it will be a wonderful computing resource.

E

For the over 8 000 office of science, researchers who use nurse it's been designed to support.

E

Not only scientific simulation, but also data, analysis and artificial intelligence applications.

E

The portion of perlmutter that is being dedicated today is composed of over 1500 nodes. Each.

E

Containing four nvidia a10 tensor, core grapple graphical processing units and one amd milan.

E

Cpu processor, because of this heterogeneity pearl motor, will provide an excellent bridge.

E

For the nurse users who want to use the exascale machines that will be deployed.

E

Over the next two years, at oak ridge in argonne, national laboratory, the lab and nurse are to be.

E

Commended for their efforts over the past year to complete the preparations necessary for the.

E

Deployment of promoter, which included electrical and cooling upgrades and for building a networking.

E

And monitoring infrastructure- this was not an easy task, given the challenges of covid19.

E

I expect to see great scientific accomplishments from pullmutter, but I want to take just a minute.

E

And highlight the dedication that the nurse staff has to its users. Nurse recognizes that preparing.

E

Applications to effectively use the system to accelerate users. Researches is an important of a.

E

Of a task as deploying the hardware as a result, in 2014, it created the nurse exascale science.

E

Application program nisap, which is a collaborative effort where nurse staff part staff partners with.

E

Application team, vendors application code teams, vendors library and tool developers.

E

To prepare for the new architectures and at that time nurse was preparing its users for corey.

E

Today there are 58 knee sap projects in place that are designed to support research on the pulmonary.

E

System these projects span three focus areas which include knee sap for senior simulations.

E

Which would be expected to prepare cutting edge simulations of complex physical phenomena.

E

And because of the breadth of promoter, they've added two additional focus areas: nisap for data.

E

To develop data analysis, science, pipelines or workflows to process massive data sets from.

E

Experimental and observational science facilities and that is joined with east nisab for learning.

E

To accelerate scientific machine scientific discovery through the incorporation of machine.

E

And deep learning, solutions in high performance, computing applications or an experimental and.

E

Simulation data analysis or both over the past year, nurse staff has worked diligently.

E

To ensure that the knee sap efforts were not significantly impacted by the move to remote work.

E

Because most coast code teams were not local to berkeley. The only difference in regular.

E

Interactions from was from what location the video meeting took place moving, perhaps.

E

From the office, perhaps to a bedroom as we tr as travel was suspended, hackathons.

E

Were transformed into virtual events and the center of excellent hackathons became.

E

12-Week series of regular online meetings, rather than one intense face-to-face workshop.

E

The goal remained the same to prepare the users in their code and we will and we'll see those results.

E

In the coming months, I have no doubt dr witherall jonathan sudeep and the entire nurse team.

E

On behalf of oscar and the nurse users, I want to thank you for your tireless efforts over the past.

E

Years to prepare for and deploy perimender the latest in a long line of nurse supercomputers.

E

Congratulations, thank you barb, and thanks for the partnership uh with oscar on this project,.

A

I'm gonna also take the opportunity now to say some thanks to our our people, who are here.

A

Representing our elected officials representative, barbara lee, senator feinstein, senator padilla.

A

Have staff members here thanks for joining us today uh and now I would like to turn to our.

A

Next speaker, who's our close partner and friend, uc berkeley's, chancellor carol, crist.

A

Carol's impressive career in university management leadership has been marked for mark by her passion.

A

For quality, accessible public higher education and her advocacy for the value of a broad education.

A

In the liberal arts and sciences and she's also been a champion of women's issues and diversity.

A

On college campuses, she started as a uc berkeley english faculty and was soon tapped for.

A

Higher service, notably, she became the executive vice chancellor in provost.

A

And she sharpened the university's intellectual focus and advanced major initiatives in areas like.

A

Neuroscience and bioengineering then she became president of smith college, one of the country's.

A

Most distinguished small liberal arts colleges and there she supervised the development of the.

A

Nation's only accredited engineering program at a women's college now back home, fortunately, for us.

A

In berkeley, as chancellor carol finds time to help us here at the lab serving on our advisory board,.

A

From her precious time and she's been a strong advocate for lab university research, partnerships.

A

Just her friendship and warmth and clay geology has been invaluable and for both our organizations.

A

And for the entire berkeley community over the last year, so please join me in welcoming.

A

Carol chris to the virtual stage good morning, and thank you mike for the introduction.

F

The relationship between the uc berkeley, campus and the berkeley lab is deep, long-standing and a.

F

Profound benefit for both institutions for our communities of scholars, teachers and students.

F

And, of course, for the public we serve, and so I'm thrilled to be part of this gathering.

F

Honoring, the birth of the newest member of our family, the pro mudder supercomputer, I'm.

F

Also honored and delighted to welcome the distinguished guests who are with.

F

Us today, including david turk, deputy secretary of the department of energy, steve binkley, principal.

F

Deputy director of the department's office of science and barbara helland associate director.

F

Of the office of science's advanced scientific computing research program, I want also to extend.

F

My warm collegial greetings and gratitude to the supercomputer's namesake professor saul perlmutter.

F

The decision by the national energy research scientific computing center to name this marvelous.

F

Machine after saul was a wise one. It was at this facility and through its computing resources that.

F

He conducted the extraordinary explorations that led to a well-deserved nobel prize.

F

It seems apt and fitting that a computer destined to expand and accelerate our understanding in an.

F

Extraordinary number of fields, everything from cosmology to materials, discovery is being named.

F

For the man who discovered that the universe itself is expanding at an accelerating rate.

F

Now more than ever, the world needs and depends on the work of institutions and individuals.

F

Who believe and trust in science who are ready and able to challenge the status quo?.

F

And who are dedicated to advancing the greater good? These are among the key traits that uc.

F

Berkeley and the berkeley lab share. We are fortunate to have the pearl mudder super.

F

Computer and if I may say so, it's fortunate to have us, I know we shall use its powers well,.

F

Thank you. Thank you, chancellor. We appreciate your attending and your very kind remarks, so um I'm.

G

Sadeep assange, I'm the director of nursk and uh I'll tell you a little bit about.

G

Promutter- and uh I just wanted- I have a couple slides to share so we're very, very.

G

Excited about uh unveiling promoter and the great science that we hope that it will enable.

G

I did want to take a moment to recognize the hard work of many many people who have contributed.

G

To this project- and so uh there are really too many to list on one slide, um uh ideally.

G

If we hadn't didn't have a pandemic, I would have gotten a picture, but a slide will have to suffice.

G

So anyway, jay sir vonson is the project director and becky totsky the project manager and we've.

G

Had really involvement of lots of people across nursk, almost every staff, member and.

G

Lots of people at the lab. I wanted to thank mike withrow for the strong.

G

Support of the laboratory, you know keeping our there's a large facility.

G

Upgrade component as part of this project and uh mike really prioritized that work.

G

uh And helped us overcome many issues over the past year and and get the facility ready to uh for.

G

The deployment of pearl mudder- um I also wanted to uh uh thank again um our our colleagues at doe.

G

uh Barb helen ben brown carol hawk and hemet patel for their strong support and then you know.

G

We've had lots of collaborations with uh with the vendors you know, these are uh um state-of-the-art.

G

Systems uh with lots of new technology, and so we've worked very closely with hpe nvidia uh and.

G

Amd on the deployment, so this this journaling actually started uh in 2015, and so it's been a.

G

Six-Year effort to get us to the point where we're at and so um we're now deploying phase one.

G

Of promoter and we'll be running the first jobs on promoter later in this webinar, and so.

G

We're very very excited about the system. um It'll represent a large leap in computing capabilities.

G

For our scientists, the the number one thing that our users complain about is not having enough time.

G

And so this will provide a lot more time for them to get their science done.

G

Barb had already mentioned um some of the details of the architecture, but it'll be.

G

Gp over 1500 nodes with uh each with four nvidia a100 gpus and one amd 64 cpu.

G

We're very excited about the fact that it has a brand new network from hpe cray called slingshot.

G

Which is ethernet compatible and it'll, make it much easier and faster to transfer data into.

G

The system from wherever it is whether it's at a scientific or experimental facility anywhere in.

G

The in the country or the world you'll be able to stream data into the system uh very, very quickly.

G

uh We're also very excited about the storage system. It's a 35 petabytes of uh all flash.

G

With quarry, we deployed a burst buffer to accelerate. I o, and in this case we've actually.

G

Made the entire file system all flash, and so the number two request from our users was faster io.

G

And so this should really provide much faster, io kind of another benefit to our user users is the.

G

The hpe cray shasta software stack, one key requirement for us was the ability to do.

G

Rolling upgrades, I know that uh maintenances aren't the most uh popular thing for our users.

G

uh But with the shasta software stack, we should be able to do rolling upgrades, so we won't need to.

G

To uh bring the entire system down uh to uh to upgrade the software, so so anyway, there's.

G

Lots, you know we're very excited about this uh phase. Two will be coming later.

G

uh This year and uh that'll have uh over 3 000 nodes uh with the amd uh cpus, and so this is a.

G

Will be a heterogeneous system, and uh so uh you know a lot of the computing will be.

G

On the gpu nodes, but the cpu nodes will actually represent.

G

As much computing power as all of corey as well, and so so, if you have a complex workflow and.

G

Part of it matches better to the cpu nodes. You'll be able to run that part of the workflow.

G

On the cpu nodes, I guess it's interesting also to note that the the cpus on from amd will be 64.

G

Cores each and uh that's actually very similar to the knights landing.

G

Chips as well, and so the hard work that you put in getting ready for the k, l partition.

G

Of quarry that'll help you uh directly as far as uh having your codes run very effectively on.

G

um On promoter- and so we have some uh comments from we're very fortunate from our great partners.

G

So we actually have the ceos of uh uh of hpe, nvidia and amd that uh uh that were sent to us.

H

On behalf of hewlett-packard enterprise, I want to congratulate the national energy research.

H

Scientific computer center at lawrence berkeley national laboratory on the launch of the new.

H

System per motor, we are incredibly proud that the hp craig ex supercomputer would further fuel the.

H

Lab's mission to accelerate critical breakthroughs about our climate and environment that will.

H

Improve energy resources today is a great day for science on behalf of all of us at nvidia.

I

I want to congratulate you, nursk the berkeley lab and the department of energy's office of science.

I

Promoter is a modern day marvel the world's fastest ai supercomputer.

I

With more than 6000 nvidia a100 tensor core gpus pro mutter can reach nearly four exit.

I

Flaps of ai performance, a level of capability that was unthinkable just a few years ago,.

I

Promoter's ability to fuse ai and high performance simulation will lead to breakthroughs in a broad.

I

Range of fields from material science and quantum physics to climate projections, biological research.

I

And more and while the work will start in the berkeley hills, it will go on to touch the lives.

I

Of hundreds of millions of people around the globe, all of us at nvidia are so proud to contribute to.

I

This incredible journey that began 90 years ago, with the creation of the berkeley lab.

I

So let me congratulate you again on today's commissioning of pearlmutter on berkeley labs.

I

90Th anniversary and all that the next 90 years will bring thank you to the lawrence berkeley.

J

National laboratory and national energy research scientific computing center for having amd at this.

J

Important event, we are so thrilled that our amd epic processors were selected to power. The pearl.

J

Motor system and will play a key role in advancing the research capabilities at berkeley lab.

J

At amd, we're focused on pushing the envelope in high performance computing.

J

Over the last several years, we've invested heavily to deliver leadership. Computing products like our.

J

Third gen epic processors at the heart of the perlmutter supercomputer by more than tripling.

J

The computational capacity available at nursk this system will help researchers working at.

J

The forefront of energy and materials science solve some of the world's biggest challenges.

J

We have collaborated closely over the last several years with the department of energy.

J

Lawrence berkeley, national lab nurse and hpe to bring the power of pearl mudder to researchers.

J

And we're so excited for the role we played in helping to reach today's important milestone.

J

Congratulations to everybody involved in today's launch and I look forward to the groundbreaking.

J

Research that promoter will help enable, so we wanted to thank our industry partners for the.

G

Messages and really the great collaborations that we've had over a number of years.

G

Now it is with great pleasure that I now introduce you to the 2011 nobel laureate and the namesake.

G

Of our next generation super computer uh saul permuter, as some of you know here at nurse, we.

G

Have a tradition of naming our systems after notable scientists to underscore our mission.

G

Of facilitating scientific discovery, but in our 47-year history this is the first time we've named.

G

The system after one of our users and it's really fitting that this honor goes to dr perlmutter.

G

In the late 1990s, dr prometer discovered that the universe is expanding at an accelerating rate. He.

G

Confirmed his observational conclusions by running thousands of simulations on nurse supercomputers.

G

It's believed that his research team was among the very first to analyze and validate.

G

Observational data in cosmology as a result of this groundbreaking work, dr promoter was.

G

Awarded the 2011 nobel prize in physics, his melding of data analysis and cosmology.

G

Make it appropriate that we name nursk9 promoter since it's our first.

G

Computer design from the very beginning to meet the needs of both simulation and data analysis.

G

Today, dr promutter is a professor of physics at the university of california berkeley, a senior.

G

Scientist at berkeley lab and still an active nurse user as the director of the burst play.

G

Institute for data science, he promotes scientific breakthroughs by advancing interdisciplinary.

G

Data driven discovery. He is also executive director of the berkeley center for cosmology.

G

Cosmological physics and leads the international supernova cosmology project. He also leads an.

G

Interdisciplinary education team at uc, berkeley that teaches a scientific style of critical.

G

Thinking for undergraduates in a class called sense, sensibility and science. Please welcome.

G

Me join me in welcoming dr saul poormunder, and I just wanted to say that when I first asked.

G

Saul if we could name the system after him, one of his first comments was: are you going to make.

G

People type in promoter- and so I just wanted to let our users know that you will have the option.

G

Of when you ssh in you can ssh into promoter promoter.nurse.gov.

G

Or to saul.nurse.gov in case that's uh easier, so thank you. Thank you, sadie, and, and thank you to.

K

All the distinguished guests today, I'm I'm very honored by this occasion and and this naming.

K

Let me begin by saying that um I I also felt very honored when sleep first called me and told me.

K

That they wanted to name this computer promoter. And but I should also say that, when steve called.

K

I was a little worried that this might be a bad idea and I actually had two reasons. um First,.

K

Was the one that suddenly is referring to that after spending a lifetime of trying to type my.

K

Own name quickly and spelt correctly now this was potentially dooming a whole generation of.

K

Nurse supercomputer users to perform the same feat whenever they logged on. So that was on the one.

K

Hand while, on the other hand, I was having visions of all those times when people will be discussing.

K

How promoter is down for upgrades this weekend, um although now I hear that they there will be.

K

uh Rolling upgrades or um or when the computer becomes so popular that people start complaining.

K

That promoter's q is running slow today, um but as you as you've heard city promised me.

K

That they could have an alternative, alias uh using my first name, so people would only have to type.

K

Four letters, um but even more important in the end, was able to make me feel um feel uh better.

K

Good about the whole thing by explaining that they really wanted to use my name.

K

Because they wanted the name of the computer to represent that move towards more data intensive.

K

Use of super computers- um that's you know the a little more. The experiment focused work.

K

As opposed to the ever popular simulation and theory, work, and, and also they wanted the name to.

K

Represent the sort of group based efforts um that I had always done so, for example, they would use a.

K

Picture of the whole group on the computer console um you know when, when the graphics were set up.

K

So those these points made me feel more excited about the the naming idea and, and it's and it's.

K

Actually true, that nurse has been a a big part of my own research life so from the from the very.

K

First day that it moved to lvnl years ago, just uh then just a few flights downstairs in building 50.

K

And I I always find myself waiting for each of the new generations of supercomputers.

K

And and using them for our work it it made new things possible that weren't before.

K

And now we're actually living in a period in which data sets are getting still bigger and we've also.

K

Developed new ways of analyzing statistics that need multiple simulations of the entire data sets.

K

And so both of these are calling for this new super computer um and furthermore, this is really.

K

A team sport, um both the data collection and nowadays, that's the all. This analysis.

K

These are all being done by teams, as, as was all the work I've done.

K

um So I'm really right now eagerly waiting, not not only the chance to use the computer myself, but the.

K

Chance to see what today's young postdoc, who you know, who's my equivalent just a few decades later.

K

Will will think up to use it for and I'm it's just going to be wonderful to see what the.

K

Next berkeley, postdoc and of that matter, the next postdoc you know anywhere in the country.

K

Who is now networked into nurse gets to do. Finally, then, let me just congratulate the the whole team.

K

At nursk and its partners, and that worked on the huge job of getting this, this new.

K

Computer designed built and installed, and up and running and my anticipatory congratulations to.

K

The next generations of scientists who will be discovering all sorts of great things with it.

K

So, thank you again and let me pass the ball back back to sudhi.

G

Okay, thank you. Thank you saul for your uh your your great comments and we're very excited.

G

And uh saul will come on later and actually launch the first job on uh promoter as well. So.

G

Something to look forward to uh now. I have the honor of introducing my long time friend.

G

And colleague horse simon, an internationally recognized expert in computer science and applied.

G

Mathematics and berkeley labs deputy director for research, dr simon, will moderate today's panel.

G

Discussion about how the perlmutter system fits in the to this moment in hpc, and what we.

G

Can expect of its impact on science and computing technologies over the next five years and beyond.

G

He joined berkeley lab in 1996 as the director of nurse and played a key role in establishing.

G

The facility at berkeley, under his leadership nurse, became a world leader in providing super.

G

Computing resources to support research across a spectrum of scientific disciplines.

G

This integration empowered many important discoveries, including dr prometer's, noble.

G

Prize winning work and it's making it possible for us to sustain a leading role in the transformation.

G

Of science that is resulting from deeper applications of computing and mathematics.

G

Dr simon was also a founding director of berkeley labs, computational research.

G

Division which conducts applied research and development in computer science, computational.

G

Science and applied mathematics as associate laboratory director for computing sciences.

G

Area he worked to foster interdisciplinary collaborations among researchers at berkeley lab.

G

And between the lab and uc berkeley in the hpc field, dr simon is known for his algorithmic.

G

Research efforts and was honored with the gordon pell gordon bell prize in 1988 and 2009 for his.

G

Parallel processing research, he is also co-editor of the bi-annual top 500 list that tracks the.

G

Most powerful supercomputers, worldwide and related architecture and technology trends. Please join me.

G

In welcoming horst to the virtual stage, it's a horse, I I remember when we met uh in the early.

G

90S and I guess we looked a little bit different back then, but it's really been my pleasure to.

G

Been your colleague for uh an extended period of time. Thank you cynthia for this very generous.

L

Introduction- and it's obviously a great pleasure to be here next to you to and join you for the.

L

Nurse 9 promoter dedication, as you've said, I've been directly indirectly associated with nurse for.

L

More than 25 years and as a nurse director from 96 to 2006, so it's a great pleasure to share a.

L

Forward-Looking panel, but in order to appreciate the potential impact, a new system such as.

L

Promoter might have over next five years. I want to briefly look back over the last 30 years.

L

Go back all the way to nurse one which was a crazy 90 with about 13 gigaflops performance.

L

And after a steady series of new systems about every three to four years today, we are looking.

L

Forward to dedicating nurse 9, which on back of the envelope, is about 7 million times more.

L

Powerful than nurse 1., this is a tremendous accomplishment and this huge performance gain.

L

Has, as was discussed, also translated into comparable scientific accomplishments.

L

So, looking back today, I'd like to thank the ue office of science and oscar.

L

uh Represented by steve, binkley and barb helen here for their consistent support and planning.

L

Because, without that support and planning, we would not be here.

L

I also want to thank the nurse staff that has been very dedicated over the years, introducing.

L

New technology maturing it and working with users to make it a useful tool for science.

L

And I would like to thank the nurse scientific community that has been a long time, supportive.

L

Element of our work and has been coming up with creative things: how to use these supercomputers.

L

For the benefit of us all, so one of those key accomplishments was, of course, the analysis of data.

L

From the super mover complex module project by sole perimeter that we're celebrating here today.

L

So, with this background, the huge performance improvement, the scientific impact I want to turn.

L

To our panel and ask them about their expectations from nurse 9, the promoter system in the next.

L

Five years so I'd like to ask my panelists to join me on the screen by turning their cameras on and.

L

While they be coming online, I would like to invite the audience to post any questions for the panel.

L

In the chat and if we have time at the end, I will direct these questions at the panelists.

L

I want to pose a question to every one of my panelists and I'd like to start out with.

L

Ruby leong and the bio for the panelists will be posted also in the chat. So we.

L

And keep this short, probably only is a scientist at pacific northwest national lab.

L

And she is a climate scientist, and so my question for ruby is, as a climate scientist.

L

With huge challenges that we have ahead in dealing with climate change, what do you expect to be able.

L

To accomplish in your research on promoting the next five years and what you generally expect neo.

L

Exoskill hpc systems can contribute to addressing the challenges of climate science. Thank you horse.

M

um I'm going to discuss two huge challenges that we expect to address using pearl mother. The first.

M

Huge challenge in climate science is to provide more accurate and spatially detailed climate.

M

Information to support infrastructure planning and resource management under climate change.

M

Climate modeling is at the important juncture in addressing this challenge. We know that global.

M

Climate models are typically run at resolutions of about 100 kilometer, which is really far too cost.

M

To resolve clouds and convection that dominate the climate response to greenhouse gases, so this.

M

Has contributed to large uncertainties in climate simulations, the cloud resolving climate modeling.

M

Is now feasible with near exoskelet hpc systems like pearl motor, for example, does energy exoskelet.

M

Earth system model e3 exam has two approaches for cloud: resolving climate modeling that are ready.

M

To go on gpu machines in the first approach, nu3sm can be used to run simulations at kilometer scale.

M

Globally or regionally using mesh refinement for regions of interest such as the united states.

M

A recent global e3sm simulation at three kilometer resolution shows significant.

M

Improvements in modeling clouds, precipitation and extreme weather events. The second approach is a.

M

Multi-Scale modeling framework, in which a kilometer-scale cloud resolving model.

M

Is embedded within each course resolution u3sm grid to better resolve clouds, so per mother can.

M

Provide computing resources for more extensive evaluation and calibration of these approaches.

M

But even with exoskilled computers, it will not be possible to run traditional climate change.

M

Experiments that require hundreds of years of simulations using cloud resolving climate model.

M

Therefore, it is very important that a broader community of users have access to computing.

M

Resources like perlmutter to develop novel experiments and analysis using cloud resolving.

M

Climate models for climate change studies; the second huge challenge in climate science- that.

M

I'm going to talk about is the need to provide uncertainty, estimate on climate change, information.

M

That we provide this can be achieved by running large examples of climate simulations to provide.

M

Probabilistic information in phase two pro model with both gpu and cpu nodes will be an.

M

Important computing resource for large ensemble modeling using model configurations at more.

M

Standard resolutions between 25 kilometer to 100 kilometer different examples can be generated by.

M

Perturbing, the initial conditions or perturbing the model representations of physical processes.

M

Preserving the model resolution or perturbing the social economics pathway, such as options.

M

For decarbonization for the future projections, the computing resources.

M

Are needed to more fully explore different approaches for generating the large ensembles.

M

Last but not least, I I'd like to note that we have seen an increasing use of machine learning.

M

Or artificial intelligence in climate modeling, which offers alternative approaches to addressing.

M

The challenges that I just discussed, for example, machine learning, can be used to build better.

M

Parameterizations of physical processes to improve model calibration and to quantify uncertainty.

M

The pearl model will be an important computing resource for advancing the use of machine learning.

M

In earth system modeling, thank you. Thank you, ruby for your predictions, so I hope that all our norse.

L

Team listens and will be able to help you to make it all come true. Thank you now, I'd like to welcome.

L

Professor tom miller he's a professor at caltech again, his bio will be posted and I'd like to ask.

L

Tom, as a scientist who is modeling new materials for energy applications, what do you expect.

L

To be able to accomplish in your research on promoting the next five years, in particular, what.

L

Do you expect the neo accessible hpc systems can contribute to addressing materials challenges in.

L

Particular for energy applications. Thank you horst and thanks to the entire.

N

Organization for this wonderful chance to discuss the new, the new pro motor resource, I begin by.

N

Emphasizing that new materials are urgently needed for designing better energy technologies,.

N

And resources we need more efficient and scalable ways of converting solar energy. We need longer.

N

Lasting and lighter weight battery technologies and we want stronger and lighter technologies.

N

For wind and other mechanical applications and computational predictions like those.

N

Executed on perlmutter provide a valuable way to to guide those discoveries and developments.

N

But designing energy materials is a matter of exploring and navigating chemical space.

N

The realm of possible molecules and materials, and this chemical space is just extraordinarily vast.

N

Dictated by the combinatorially large number of ways to combine atoms into molecules and materials.

N

The possibilities are truly astronomical in scale to navigate this vast chemical space.

N

We really need two things: we need the ability to accurately and quickly predict the properties of.

N

Candidate energy materials, on the basis of physical principles, like quantum mechanics.

N

But we also need a way to take advantage of vast troves of materials data previously computed.

N

Calculations previously obtained experiments for the development and design of machine.

N

Learning and ai strategies that can employ that data to accelerate additional discovery.

N

Perlmutter addresses both of these needs. Promuter will provide extremely powerful processors of.

N

Course that are faster that have more memory and they're available in greater number than.

N

Ever before this will be a game-changing resource and our ability to design better predictive tools.

N

Both physics-based and data-centric ai centric, through deep learning strategies.

N

For the discovery of new energy materials, I can speak on on in terms of my own research.

N

That there's just tremendous opportunity here that we're excited to take advantage of in.

N

Terms of the pro model resource- and I know that it's going to just cause a an inflection point.

N

In terms of the the capabilities that are available to the community.

N

Thank you thank you tom, and thank you for your vision of integrating the data resources with.

L

The computing resources, which maps perfectly in what we're trying to accomplish with kilometer.

L

Next, I would like to ask I'd like to welcome kathy ellick, who is of course well known since.

L

She was the nurse director from 2006 to 2012, I believe, and kathy ben was the associate lab.

L

Director in computing sciences, but she is now has a new role as associate dean for computing.

L

Data science and society at uc, berkeley, but kathy, is still a user on nurse actively involved.

L

Some biological applications so I'd like to ask you: what do you expect the role of hpc platforms.

L

Such as promoters will have in data science and project and machine learning, which two of.

L

The previous speakers have already commented on that. I know you have worked in this direction for.

L

The last multiple years, so, thank you. Thank you very much first and uh thank you for inviting.

O

Me to speak to all of you, I'm very excited about the incoming pearl mudder system. In my role as.

O

An associate dean for the for research in the division of computing data, science and society.

O

Which is a unique academic structure at uc berkeley that combines the school of information.

O

The department of statistics and the department of electoral engineering and computer sciences.

O

Shared with the college of engineering, along with cross-disciplinary structures to hold.

O

Various types of education and research programs perlmutter is going to be a tremendous resource.

O

I think for this division, which we also have trouble in terms of the length of the name, so we.

O

Refer to it as cdss, and um so we're very excited about what perlmutter can be used for, including.

O

Simulation problems, traditional data analytics problems and these machine learning applications.

O

And ai techniques, more broadly so, and I also want to just acknowledge that, in addition to research.

O

That nurse has served the educational mission of the university I've personally used it in.

O

My classes for teaching over a thousand graduate students and undergraduate students over the last.

O

Decade to teach them about high performance parallel computing and to use the kind of.

O

Systems that that nurse has provided. So, in addition to, though my role as associate dean.

O

I'm also a computer scientist and I could spend some time geeking out on the technology in pearl.

O

Mudder and I'm very excited about the combination of the latest gpus that you've heard about.

O

As well as the high-speed interconnect, that's going to make it a unique resource, and especially.

O

I think a unique resource for running the kinds of clever and efficient computationally efficient.

O

Algorithms, that can be run at unprecedented scales on this on the perlmutter system. So a little bit.

O

More about cdss and how I think we may use pearl mudder. We I want to first of all acknowledge one.

O

Of the cornerstones of the cdss research program, which is the berkeley institute for data science,.

O

Bids is led by none other than saul perlmutter, and um so, in addition to lending his name to the.

O

Nurse system, I also want to thank him for his close partnerships in developing the research.

O

Initiatives within cbss, those cross-disciplinary initiatives run the gamut from fundamental.

O

Problems in cosmology to applied problems and energy efficiency, but there are three.

O

Top initiatives that I want to talk about that, I think, will be for which systems like perlmutter.

O

Will be critical and in particular use of machine learning techniques will be important, so the first.

O

One is in climate change which you've already heard about, but and we're focused on some of the.

O

Machine learning and data intensive aspects of this, in addition to the simulation problems.

O

That ruby spoke about, but these fall into both mitigation and adaptation, so things like using.

O

Machine learning models to drive the design of materials that can be used for carbon capture.

O

To remove carbon from the atmosphere, also using machine learning to analyze, satellite imagery.

O

To build reusable models which has been done so using incredible amounts of computing to.

O

Train these models and those models, then, can be used for a diverse set of application. Questions.

O

That arise such as looking at forest cover to what is the level of poverty in various regions.

O

And the second initiative looks at health applications and that intersects with nurse's.

O

Mission in some of the fundamental biological questions that arise in synthetic biology.

O

And in the environment, as well as in health, things like using machine learning methods, including.

O

Graft neural nets to understand what proteins do to discover a new protein when you're looking for.

O

Something that has a particular characteristic and then figuring out how to manufacture that protein.

O

At scale and use them to use biological processes and these kinds of inverse design problems can be.

O

Used to engineer biological systems that produce everything from biofuels to pharmaceuticals,.

O

Graft neural networks, as I mentioned, one of the techniques that may come up in understanding these.

O

Protein structures can also that will also take advantage. I think, of the unique gpus as well as.

O

The high speed network of perlmutter they're, very computationally, intensive and have very irregular.

O

Sorts of patterns of of communication and memory accesses our last initiative and cdss is in human.

O

Welfare and social justice, and while cdss is looking at things like the opportunities and the.

O

Risks of using algorithmic decision making, such as machine learning, algorithms in criminal justice in.

O

Distributing social welfare support and in public health, I think within the department of energy.

O

We'll be looking at things like energy justice and how the deployment of advanced energy.

O

Infrastructure can be used to empower communities to realize their ambitions, and this brings up a.

O

Set of core problems in cdss in that those those fundamental questions around algorithmic fairness.

O

Around privacy and computing on private and encrypted data and on trust in data and bias.

O

In data- and I think all of those questions will they arise in both fundamental science questions.

O

And uh questions in energy applications, as well as in this broad set of applications that we're.

O

Looking at in cdss- and finally I just want to mention that I think doe in addition to.

O

Focusing on the office of science mission that nurse has has also been deployed in in important.

O

uh National and international problems, including the kobandain teen pandemic,.

O

And with the support of people like barb, helland and steve binkley, it was brought to bear um in the.

O

Incovid 19, the nurse resources were as part of the hpc consortium and the the resources across all 17.

O

Labs as the part of the national biotechnology laboratory and I'm sure, pearl mudder can be.

O

Viewed as both the unique resources for these broader challenges in the future, as well as.

O

In in its kind of bread and butter work, uh solving mission problems in the office of science.

L

Thank you kathy for a very broad range of potential applications in the future. All of them.

L

Very exciting and I'm looking forward to see you.

L

Bringing more of the uc berkeley students into the hpc world and explore all these challenges.

L

So now I'm very happy to welcome irene qualters who joined us after.

L

Some technical difficulties just in time welcome irene, happy to see you and I'd like to.

L

Thank her for joining us irene has a long experience in the computing industry than.

L

In government working at the national science foundation she's now the associate lab director.

L

For computing at los alamos national lab, so with this broad environment experience, how do you.

L

Expect the hpc environment to develop over the next five years. What will be the big challenges.

L

Ahead for high performance computing and for large centers such as nurse, thank you horst and.

P

I I wanted to actually start back up a little bit and add my congratulations, and also my thanks from.

P

Los alamos national lab to doe office of science, berkeley lab and nurse for this very well-timed.

P

Forward-Looking leadership, contribution to hpc science community and so as sudeep illustrated.

P

With your collaborators, you've really embraced full force. The heterogeneity.

P

That is coming that we're just on the leading edge of and so um you've brought about novel.

P

Architecture, I'm not going to repeat what su deep mentioned in the novel technologies in the system.

P

Barb also talked a good deal about the knee sap, the application focus, but I think that.

P

All positions, the scientific communities that uh kathy and others have spoken about to advance to.

P

The next frontier, and so and then the third thing I wanted to call out, is by focusing on those three.

P

Primary modes of computing simulation uh data and learning you are really positioning the community.

P

As a whole for the future, so now a little bit about what I see as the challenges, I.

P

I think we're really in the leading edge of exploring and trying to optimally navigate.

P

A rich and disruptive- and I know that word is often used, but I think it's true of semi-custom.

P

Heterogeneity in the technology basis for science and- and so I I think one of the challenges will be.

P

How we continue to really uh rapidly take advantage of these new technologies.

P

And the new methods, the algorithms and the product approaches to prediction, control.

P

And decision support, while we sustain the underlying foundational intellectual capital.

P

That's represented in the applications, so I I think how we do that will really.

P

Allow us to either advance rapidly or get stuck a little bit in in trying to build.

P

On what we've already achieved, and then the um the second piece is a little more technical.

P

One sees evidence in shasta, its use of containers and in other technologies of hpc being.

P

Less isolated than in the past and being a vibrant, contributor and benefactor of the.

P

Expanding technology, landscape of sensors, data's instruments workflows, but I think being able to.

P

Optimally tap. That is really going to be a challenge uh for us all. So that's that's all.

L

Thank you irene, and I would like to now turn over to sudeep, who has already been introduced, and the.

L

Question for the director of nurse is since very simple, but very difficult to answer the simple.

L

Question is what are your hopes and expectations for promoter and for next and for nursing in.

L

The next five years well, thank you horace and thank you for inviting me to be on the panel.

G

um You know it's an exciting time. You know, there's a considerable jump in.

G

Computing power that will hopefully drive new discoveries in many scientific fields such as.

G

Cosmology, clean energy, climate and the environment, microelectronics, materials, science to name a few.

G

In addition, as I mentioned before, promoter is the first system at nurse to be designed to.

G

Meet the needs of simulation data analysis and ai at the same time, and really most.

G

If not all scientific fields are facing a deluge of data, light sources, accelerators telescopes,.

G

Sensors sequencers are generating more and more information, and you really what's changed. Is you.

G

Really need a super computer to analyze that much data. You know right now. More data comes into nurse.

G

And leaves nursk, which is really a paradigm shift for a super computing center and io has become.

G

Even more important than before, not only io to the scratch file system, but also data transfer.

G

Over es net to remote scientific instruments, and so so really you know, I think, we're going.

G

To see a broadening of the uh the user base at nursk, you know we already have 8000 users.

G

I imagine that that's going to even go up as we start to serve these new communities.

G

And you know, I did want to note that our work won't be done when promoter goes into production.

G

uh You know nice, the nissan teams have worked really hard to get ready for the gpu nodes.

G

However, we really want to make sure that the very broad user base that we have you.

G

Know with 8 000 users and 700 codes that they can really use the gpo gpus.

G

Nodes as much as possible, and so we'll continue to work on training and helping the code teams.

G

You know in in addition, much like the early days of parallel computing.

G

You know with horse that you and I were involved with- we really need to tackle the hard problems.

G

The things that don't easily go on gpus to really to really broaden the impact of this technology.

G

uh You know over the next five years. Actually, even now, we started to plan already for nurse 10 and.

G

Nurse 11., you know these systems uh takes years and years for us to uh to design and uh deploy.

G

um And it's you know pretty clear that we're going to have more heterogeneous systems as we enter the.

G

The postmoore's law era, you know we're looking at different types of accelerators. You know I don't.

G

Think that it's likely that nurse 10 will have a quantum accelerator, but nurse 11's certainly might.

G

uh Half the codes that run at nurse uh solve some kind of quantum mechanical problem, and so.

G

Those might that part of the workload might actually really benefit from a.

G

uh From a quantum accelerator- and so um you know with nurse 10- we're really going to focus.

G

On end-to-end doe office of science workflows- and you know, hopefully enable new modes of.

G

Scientific discovery through the integration of experiment, data, analysis and simulation, and so not.

G

Only do we want to make sure that the scientists can use ai to analyze their data, but we also want.

G

To use ai to manage the system to increase the reliability of the system and the energy.

G

Efficiency of the system and in addition, we have a goal of using ai to to reconfigure the nurse 10.

G

To accelerate workflows in real time, and so so I think it's a very exciting time. We're going to see.

G

You know lots of uh lots of great new technology. You know we, you all have my commitment that will.

G

Work with the nurse users to make sure that we help them so that they can use this technology.

G

As easily as possible and maximize the impact it has on their uh their science, so thank you.

L

Thank you so deep and you answered a wonderful questions. That was a question about the role of.

L

Quantum computing, so thank you for thinking ahead. I have uh two quick questions that I think you can.

L

Answer very quickly will that's for so deep will corey be around for some time or will perimeter.

L

Retire glory in the near future. What are the quarry plants yeah? So so yeah don't worry.

G

Cory is not going to go away uh soon, so we'll be running corey at least through 2022, and so you.

G

Know we'll make that decision uh in collaboration with our funding agency at uh at oscar. You know.

G

With uh ben brown and and carol hawk and and barb helen, but you know we would not, uh we would not.

G

Turn off uh uh cory until um perlmutter is, uh is, is very stable and producing uh lots of great.

G

Science now uh promoter will be much more powerful than than corey, and uh so you know at some point.

G

um That added computing power will not be, as you know, as impactful as it is right now. Okay,.

L

Thanks and then the second one, that's a very simple one. uh What is the theoretical flops of.

L

The phase one of the motor and are you planning to submit an hpl benchmark result. I didn't with this.

L

Somebody else said yeah, so the yeah, so the the the peak speed is about uh 120 petaflops of phase one.

G

uh You know we really optimize the system for science and our users and not for.

G

Our position on the top 500 list, and so if we just wanted to get the top position on.

G

The top 500 list we wouldn't be deploying a cpu partition, for instance, and so really.

G

You know that's our goal, but hpl is a great test of a system. It's it's reliability.

G

And the fact that you know can all all the nodes can work together, and so we will be running hpl.

G

The uh the the peak currently it's about uh uh pedal flops per cabinet and we have 12 cabinets.

L

Thank you. um I have another question that I'd like to ask, and hopefully all the other panelists can.

L

Moment in this one as the amount of data from experimental systems and facilities continues.

L

To grow exponentially, can we maintain the current schedule of upgrading the eu hpc systems.

L

This nurse can lead to growing capabilities and productively every two years. Maybe every.

L

Year, this is even possible and can math and software advances help us to keep up.

L

With data, so I think it's a two separate question, so does anybody want to comment on whether.

L

Math and software will help us to keep up with the data explosion. I can chime in on that.

N

Thank you with regard to the opportunities for from athens for advances to keep up with the data.

N

um I think that's kind of a central aspect of of representation learning in in in machine learning.

N

Strategies finding ways to take things and find connectivity among them find structure among them.

N

And that can lead to compression and storage strategies and organizational strategies for.

N

Databasing as well, so I think this task of of uh managing the explosion of data is.

N

Intrinsically connected to the algorithmic developments on the machine learning side.

O

Okay- maybe I'll just add on the the other point about keeping up. I think that um I they do think.

O

That the dui systems um will, you know, need to deal with the effect that there's experimental.

O

Facilities within the office of science that have these exponentially growing data rates.

O

And that has not traditionally been the role of the hpc centers, but I think increasingly they.

O

Need to be able to support them, and that brings in some unique sorts of challenges that I know norsk.

O

Is working on in terms of being able to integrate with the facilities and get co-scheduling, and so.

O

On but I think the other thing which wasn't really part of the question is and that I've seen very.

O

Much on the berkeley campus is when you are doing research on machine learning. The computational.

O

Requirements that you have are far exceed what those same researchers needed. You know.

O

Five or ten years ago, and so I think that we have not really come to grips as a country.

O

um The amount of computing that we need to deploy for the research community.

O

For ai and broadly for machine learning, applications across science.

P

Maybe I could just add to that. I do agree that we have not yet come to grips with that. on the.

P

Other hand, there there is the um trade-off between um uh using a platform and optimizing.

P

And learning from that and using that to feed in to the future, and that takes some time.

P

um And so you know the the cadence uh there's a certain cadence that could be upset within every.

P

Two year, where you actually don't get the return, because you're always on the move, if you will.

L

Okay, well, I think at this point I would like to.

L

Thank all the panelists for joining me on the virtual stage and for sharing their.

L

Vision for the next five years, so those of you who are using the system- I wish you all.

L

Best of success with your scientific and viewers and the ones who are managing the system.

L

Like irene and sadiva, wish all the best of luck in maturing for technology and making it useful.

L

For the uni mission, I would like to thank all the panelists for joining me on the virtual stage and.

L

For sharing their vision for the next five years, so those of you who are using the system, I wish.

L

You all best of success with your scientific and viewers and the ones who are managing the system.

L

Like irene and sudeep, I wish all the best of luck in maturing for technology and making it useful.

L

For the ue mission with that I'd like to ask my panelists to to say farewell, and I would like to.

L

Turn this over now to the highlight and key event and welcome the doc jacobson on stage and doug.

L

Is a group lead for the computational systems group at nursery this good team that takes.

L

Care of promoting will make sure that it will be operational and performing well.

L

And so uh the teams at nurse hp have been an amazing job at bringing together the promoter.

L

Super computer so we're confident with the result that doug and his vendor partners have.

L

Accomplished where we're going live now to the opening of promoter, and so I think it's all please.

L

Join us on camera and I will fade away duck go ahead. Yes, hi, I'm I'm doug and uh yes thank you for.

Q

The wonderful um introduction uh we we have been working very hard on this super computer and it's.

Q

You know doing this during the pandemic has been quite an experience. We're gonna do a live demo.

Q

Or sorry opening right now, dr promoter, would you please join me? It's a pleasure. So, dr promoter, we.

Q

Would be honored if you would open the system to run its first scientific calculations, I'm going to.

Q

Share my screen now to walk us all through these very special on the left side of my screen. You can.

Q

See the text user interface, the scientist used to submit calculations to the promoter supercomputer.

Q

You can see the interface scientists use to review and observe what calculations are running on the.

Q

System at any, given time once you open the system doctor promoter, the right screen should populate.

Q

With running calculations, soon after I'm going to share control of my screen with you now, okay, so dr.

Q

Prometer we've granted you a bit of administrative control of the supercomputer. For today's event,.

Q

Would you please click into the terminal on the left and then please enter the command.

Q

To open up the system there, it is right, okay, so then, dr promoter, after the countdown from five.

Q

Please press turn to allow the first official scientific calculations to start running on the.

Q

Promoter supercomputer everyone, please join me as we count down five four three, two one.

L

There, yes, and we should start seeing the job spot there. They are. Okay, that's excellent, so thank you.

Q

So much dr promoter, the list of calculations is now appearing on the right side of my screen.

Q

As they begin, you can see each one represented as a line of the sq output on the right screen.

Q

And you know we can see, there's a large diversity. These are using the gpus.

Q

That's great, um I can't see everything through my own screen. That's fine! Oh well! There's even more.

Q

Okay, it looks like that's all coming through, so let's switch to an even more colorful display.

Q

That has been prepared- and this is also live data, and we can see that each of the different projects.

Q

That have been uh selected to run for today is is coming through now. So that's that's.

Q

Phenomenal, it looks like the machine is doing exactly as it's meant to do. That's wonderful.

Q

um Okay, okay, so now I'd like to invite the whole nurse.

Q

Team to join me in celebrating this moment. uh Congratulations, everybody.

Q

Thank you very much, dr promoter for this today. Congratulations.

A

And thanks doug and thanks all that was great- that was great, uh very impressive, very inspiring.

A

uh I want to thank uh everybody on the nurse team for doing uh putting in all the work to bring this.

A

Us to this point, giving such a memorable beginning to the life of the perlmutter.

A

Supercomputer, thanks to our speakers, who have made this event even more meaningful.

A

Thanks to the teams from the computing sciences area, strategic communications and protocol for.

A

Putting together such a terrific event, thanks for all the people, who've been working hard for the.

A

Last year, to make this possible, and finally I'd like to thank my physics colleague, saul promoter,.

A

For letting us name our system after him in his nobel prize-winning work, he not only has a road.

A

On the hill, he also has a computer. I hope that all of you enjoyed this event as much as all of.

A

As I have, and I think all of us have it's been a pleasure to celebrate this momentous occasion.

A

We look forward to seeing you all in person soon.