(music)
>> Hi, this is
Steve Michelotti of
the Azure Government Engineering team.
I'm joined here today by Anthony Robbins from NVIDIA.
We're going to talk about GPU
on Azure Government. Welcome Anthony.
>> Steve, thanks for having me.
>> So, I think we should start out
in the most logical way which is tell me-.
Let's talk about GPUs, give us some background,
what's all the buzz about.
Why do we care about this?
>> A lot going on with GPUs and a lot of this has
been inspired by the fact that for the last 30 years,
we've been able to get a predictable
50 percent performance improvement
from microprocessors every year.
And this is because the amount of
transistors we've been able to put on
the chip has improved
such that we could get that performance.
What we have found over
the last several years is that it's trailing off.
So, instead of a 50 percent performance improvement,
we're like at 10 percent.
Now interestingly, that's happening
at the time in which the worlds networks
are getting faster and
the world's data is more and more than we've ever seen.
So, we've got faster networks,
more data, slowing microprocessor performance.
And at the same time,
what we at NVIDIA have found a way to do,
is get accelerated computing capabilities
via these things we call GPUs.
>> Okay, great. So, definitely an exciting time to be
a technologist with everything
we have going on right now.
Talk to me a little bit more about
where NVIDIA fits into this.
>> So, our GPUs,
we first brought out into
the high-end gaming and graphics marketplace.
And then what we found is if we could program these GPUs,
we could make them more
available to general purpose computing.
And so, we came out in
2006 with something that we call CUDA.
And CUDA is a programming model that allows
users to unlock the capability of GPUs.
And that began to make
GPUs more widely available in the marketplace,
and begin giving high-performance capabilities
to broad users in industries around the world.
>> Okay. That's very interesting.
Imagine lots of different scenarios
where you can be brought to bear.
I know that in Microsoft specifically A.I.,
Artificial Intelligence is very interesting to us.
We fancy ourselves one of the world leaders.
Where does NVIDIA fit into A.I. in that landscape?
>> So, not only do we consider Microsoft the leader
in this space corporately committed and the like.
So, there's a great partnership opportunity
for a few reasons.
As we have been investing in GPUs and CUDA,
somewhere around 2012 and 2013,
there began to be some breakthroughs
in the area of image recognition,
which began this kind of
bringing the world to think about how A.I.
might inspire enterprise of the future.
From 2012 on, what
we have found at NVIDIA is our hardware,
GPUs, and our software CUDA,
has been at the heart of most of
the world's progress in A. I.
So, we're delighted with a chance to team with
Microsoft in bringing
Cloud capabilities to the government.
There's a whole bunch of used cases
that make sense to Federal Government.
If we're talking about civilian agencies,
public safety is a key is a key opportunity.
Microsoft has got a ton of skills there.
We think GPUs and CUDA and the Microsoft,
Azure Cloud, Government Cloud, work really well.
But there's also others, right?
There's waste, fraud and abuse.
The U.S. spends 147 billion dollars
a year in waste, fraud and abuse.
So as taxpayers, I like to think we have
responsibility to help them with that business problem.
That business problem has at its core,
lots of data, so I think we can help there.
As well, health care is
a big focus for both of our companies.
I think we can add value to the civilian agencies and
state local government from a workload perspective.
The other thing really important is
our response to natural disasters.
The U.S. last year spent
306 billion dollars responding
mostly to hurricanes, right?
So, if we think about the role of FEMA,
or we think about the role of NOAA with
respect to weather forecasting and modeling.
Great opportunities for accelerated computing or GPUs,
CUDA, working along with Microsoft.
And then of course, if I go into
the Department of Defense,
command and control is a is a big use case.
The area of cyber.
An admiral at Cyber Command stated that
artificial intelligence is going to play
a fundamental role in the future of cyber security.
Platform sustainment in DOD is another area,
as is transportation logistics and autonomy.
The point Steve, is the use cases are many,
and the opportunities for
a world class company like Microsoft,
the Cloud capabilities that you guys have,
and then the technical merits of our product in your
Cloud offer us a chance to serve customer very well.
>> That's awesome.
Yes. Definitely, these use cases
are incredibly compelling
from mission scenarios to citizen services.
And you did a great job I think delineating
a lot of those different use cases and services.
Okay. So, we've been talking pretty big picture here.
Let's talk a little bit more specifically
about Azure Governments specifically,
and how NVIDIA fits into that.
If we look actually at the slide here,
we see that we have some of
our NC Series compute machines that we have,
GPU-powered machines and Azure Government,
I notice that we have this one here InfiniBand.
Can you talk a little bit about
what's interesting about InfiniBand.
>> Importantly, pulled out on the slide.
>> Yeah, exactly.
>> It's one of the important characteristics
of the advantage
of the Microsoft Azure Cloud to our government customers,
because it gives high performance connectivity,
which a lot of the use cases that I described,
whether it's high performance computing
or accelerated computing requires.
So it's a technical differentiator
for Microsoft and I think it's
a big one to pay attention to.
>> Definitely excited about it.
And this just gives you different skews
depending on who you are as a customer,
and what your price point is, how much power you need,
you have different options to choose from here.
Okay. And so I noticed that we also
have these visualization VMs.
What's from a high level of a differentiator
between the visualization VMs versus Compute.
>> Well, no solution from us would
be complete if we didn't offer this as well.
So at the start of
our company in the founding and where we
first brought dominance to a marketplace
was in high-end gaming and graphics.
This allows us to bring graphics and
visualization capabilities to the Cloud
and offered up and things like VDI et cetera.
>> Okay. All right, great. All right.
So, we talked from a high-level, we've talked a little
bit specifically on Azure.
I think it's time to get into a demo.
>> That's what everybody wants to see.
>> Okay, great. So what
I'm going to do here is I'm just going to
flip right over to the Azure Government portal.
We can see right at the top Microsoft Azure Government.
And I'm going to provision of VM with
GPU power on it, powered by NVIDIA.
Every time I do something like this I think to myself,
what would this take to
provision something like this in my own data center?
After requisition the hardware
and make sure I have room for it in
my data center and
make sure my operating system is good to go.
Let's contrast that with the experience you
get with Azure here.
So I just typed in the data science virtual machine,
which is one of the VMs we have which is preloaded with
a bunch of software for
artificial intelligence workloads.
That's probably a good candidate for GPU power device.
We'll just call this NVIDIA demo right here.
That's just the name of my machine, we'll select HDD,
we'll give this Admin username and password,
and we just have to confirm the password right here.
Okay. So, let's- what the heck.
Let's put this in production,
and let's just say,
Nvidia Demo as my resource group.
I can select my location here and these
are all the U.S. Government,
Azure Government data center locations.
I'll put it in the U.S. Gov Virginia.
Now this one is a little bit more interesting here.
We can select what VM types we wants,
and I can even have a slide here, minimum and maximum.
Let's scroll down to those NC series machines we
just looked at a second ago on the PowerPoint.
So, here we go.
Here's the NC6 machine, NC12.
Here's this one. 24R, this
is the one with InfiniBand technology you mentioned.
>> That's right.
>> Easy to get to, hasn't taken very long.
And contrast that to how challenging it is to
create data centers and
enterprise capabilities in government today.
>> Absolutely. Okay. So that
we can see the price right in here for each one.
I think NC6 is probably sufficient for my workload here.
I'm going to select
all the defaults here for the virtual network.
Just going to put this inside of.
It's going to see that already passed validation.
I'm good to go.
Yes, I'm happy with this.
I'm going to click "Create" and we're done.
It's going to take about five minutes behind
the scenes to provision but that's all.
Now I have GPU at my fingertips ready to go.
>> That's amazing. So one of
the great enterprise companies in
Microsoft and amazing CPU in
CUDA technology available to
our customers just like that.
>> Absolutely. Since I don't
want to wait five minutes not that that's a long time,
but I don't want to wait five
minutes as we're talking here.
I'm going to flip over to another GPU powered machine
that I already have provisioned.
Just one thing I want to point out
the device manager right here, we can see.
prove this to everybody that yes this is running
NVIDIA Tesla K80 graphics cards right there.
Now what I have here is a Jupiter notebook running
our cognitive toolkit technology
at Microsoft Jupiter notebook.
You can run Python workloads,
PY spark, spark clusters.
There's a slew of AI tools we bring to bear here.
This is an example of actually something I heard you
talking about earlier in our discussion here,
where you talked about
some image recognition and computer vision.
You referenced of course 2012 we
had breakthroughs and convolutional neural networks.
This is an example where you can
just come in and see image recognition
with flowers and can you see and NTK to GPU the train,
recognition to differentiate different types of flowers,
and this is just a really good tool to be able to come in
and train your models
and play with the data and that sort of thing.
>> And it relates exactly to
several really big use cases
across the public sector markets space.
>> Absolutely. So, this is a gamble with flowers.
But think about all the scenarios you could apply it to.
>> Exactly.
>> Okay, great. All right.
So, that's just a quick intro
to some of the tools you have
available that really benefit from being GPU-powered.
Okay. So, to wrap all this up,
kind of a call to action in terms of how does someone
want to get started with you on Azure government?
I know us right here we have this NVIDIA
Microsoft GPU POC program.
Can you tell us a little bit about that?
>> I can. It's it's
a really important opportunity for us to be in
service to our customers and public sector.
Because there is no doubt
that the public sector marketplace
is moving towards
an artificial intelligence inspired enterprise.
But they're going to need help, right.
They have one thing that we care
deeply about and that's a bunch of data.
So what we want to do to help the government,
is we do have to help train their data scientists.
But we have to expose them to
use cases thy're opportunities.
But we also have some capabilities
between Microsoft and NVIDIA,
where we can bring proof of concept capabilities as noted
here to our customers so
they can try things before they buy,
or they can explore the art of what's possible.
It's a really important opportunity and I think it's
a great engagement opportunity
for the Microsoft and NVIDIA teams.
>> Definitely. Just trying to make it as
humanly easy as possible.
Humanly as easy as possible to get started.
Try before you buy, these things you mention.
So, the links here we have are just
really easy ways to get started,
so you can start kicking the tires and
playing with this really exciting technology.
>> That's correct.
>> Okay. Well thanks for joining us.
This has been Steve Michelotti with
Anthony Robbins from NVIDIA. Thanks for watching.
For more infomation >> Mother hopes for answers in search for missing daughter - Duration: 1:45. 
For more infomation >> President Trump delays steel tariffs for US allies - Duration: 3:13. 
For more infomation >> Mueller Questions For President Trump Focus On Obstruction Of Justice - Duration: 1:35. 
For more infomation >> Live With Lee: New Location For Martine's Pastries - Duration: 2:46. 
For more infomation >> Energy FC picked to train future stars for women's soccer - Duration: 1:40. 
For more infomation >> Waynesboro Mayor thanks Augusta for help - Duration: 0:55. 
For more infomation >> Birthday cake for mom in heaven - Duration: 1:39. 
For more infomation >> Weather Update: Get Ready for Humidity - Duration: 4:24. 
For more infomation >> SOLVED!! I Need to Make Money Fast Online for Free - Duration: 3:43. 
For more infomation >> A New Home For Amya - Duration: 2:23. 
For more infomation >> Joe Louis Arena seats go up for sale - Duration: 2:22. 
Không có nhận xét nào:
Đăng nhận xét