英特爾公司Dhiraj Mallick - 計算會議 (Dhiraj Mallick, Intel | The Computing Conference)

Subtitles section Play video

>> SiliconANGLE Media presents theCUBE!
Covering the Alibaba Cloud annual conference.
Brought to you by Intel.
Now, here's John Furrier...
>> Hello everyone, welcome to exclusive coverage with SiliconANGLE, Wikibon, and theCUBE here
in Hangzhou, China for Alibaba Cloud's annual event here in Cloud City, the whole town is
a Cloud.
This is their event with developers, music festivals, and again, theCUBE coverage.
Our next guest is Dhiraj Mallick, who is the Vice President of the Data Center Group, and
the General Manager of Innovation, Pathfinding, and Architecture Group.
That's a mouthful.
Basically the CTO of the Data Center Group, trying to figure out the next big thing.
>> That's right, John.
>> Thanks for spending the time.
>> It's my pleasure.
>> We're here in China, it's-- You know in the U.S., we're looking at China, and we say
okay, the fourth largest Cloud, Alibaba Cloud?
>> Yes.
>> Going outside of Mainland China, going global.
You guys are strategic partners with them.
>> Yes.
>> They need a lot of compute, they need a lot of technology.
Is this the path that you're finding for Intel?
>> Yeah, so we've been collaborators with Alibaba for over 10 years, and we view them
as a very strategic partner.
They're one of the Super Seven, which is our top seven Cloud providers, and certainly in
China, they're a very relevant customer for many years.
We engage with them on a variety of fronts.
On the technology side, we engage with them on what their key pinpoints are, what is the
problems they want to be solving three to five years out, and then we co-develop, or
co-architect solutions with them.
>> So, I want to get your take on the event here in China, and how it relates to the global
landscape, because I, it's my first time here, and I was taken back by the booth.
I walked through Alibaba's booth, and obviously Jack Ma is inspirational.
Steve Jobs like the culture, and artistry and science coming together, but I walked
through the booth, it's almost too good to be true.
They've got Quantum Computing, a Patent Wall, they've got Hybrid Cloud, they got security,
they have IoT examples with The City Brain, a lot of great tech here at Alibaba Cloud.
>> So I think the technologies that they're investing in are very, very impressive.
Most cloud companies are probably not as far along as them, and looking at such a broad
range of technologies, the Brain Project is really exciting, because it's going to be
the Nexus of smart cities, both in China, as well as globally.
The second thing that's very interesting is their research and investments in Quantum.
While Quantum is not here today, it's certainly on the frontier, and Intel also has significant
investments in sort of unpacking where Quantum will go, and what promises it offers to address.
>> What I find interesting is that also hearing the positioning of, I kind of squint through
the positioning, they're almost talking Cloud-native, DevOps, but they have all this goodness under
the hood, and they're kind of talking IT-transitioning to Data Technology.
Everything's about data to these guys, not just collecting data, using data with software.
Now, that's really critical, because isn't that software-defined, data-driven is a hot
trend?
>> Yes, software-defined and data-driven is a very hot trend, in fact at Intel our CEO
and us all believe that we've entered the data economy, and that the explosion in data
is, and the thirst for analyzing that data to be able to drive smart business analytics
is really the key to this digital revolution.
I was reading an industry report by one of the analysts that said by 2019 there would
have been over 100 billion dollars spent on business intelligence.
And so, the real key is this data economy.
>> The intersection of things, and even industrial internet, IIot, Industrial Iot, with artificial
intelligence AI, intelligence Intel inside that word, interesting play on words--
>> Yes.
>> Is coming together, and we've covered what you guys were doing on Mobile World Congress
this year, where 5G was clearly an end-to-end architecture.
You got FPGAs, all this goodness here going on.
So that's 5G, and that's going to fuel a lot of IoT if you think of it like that way, but
now AI.
>> Yes.
>> It's Software.
How does that connect?
Because that's the path we see forward on the Wikibon analyst side, we see software
eating the world, but data eating software.
And now you got 5G creating more data.
>> Yeah, so the way we look at it at Intel is, we have data-center technologies that
are fueled by the growth at the Edge by IoT devices, because they're creating demand for
more processing capability to be able to unpack and analyze that information, and it's a self-fulfilling
circle.
We call it the virtual cycle of growth, because the data center feeds IoT demand and then
IoT feeds the data center.
And so it's the combination of those.
What 5G does, is 5G forms the connectivity fabric between the data center and the Edge.
It allows data to be pre-positioned at the correct places in the network, so that you
minimize latencies through the network, and can process or do the analytics on it as quickly
as you possibly can.
>> So we were talking before we came on camera about Jack Ma, they call him Jackie Ma here,
keynote being very inspirational, and talking moving to a new industrial era, a digital
economy, all that good stuff, very, very inspirational.
Let's translate that into the data center transformation, because we're seeing the data
center and the Cloud with Hybrid Cloud become really critical to support what you were just
talking about which is, how do you put it all together?
It sounds so easy, but it really is difficult.
>> It is, and so our vision is that in order to be able to fulfill this data economy, we
will need to have five key innovations in the data center.
The first innovation, in no particular order, is that the data center will be frictionless.
And what I mean by frictionless, is that there will be zero to low latencies in order to
provide that real-time experience at the Edge.
So latency is extremely critical, and the way we believe that that can be achieved is
by moving from copper to light.
And Intel has significant investments in leadership products and silicon photonics that will enable
switches to be based on photonics.
It'll enable CPUs, and server hosts to be based on light.
So we believe that light is a critical aspect to this success.
The second aspect of frictionless is the need for liquid cooling and that was in the keynotes
from Simon Hu this morning, that the liquid cooling is going to be essential to be able
to enable a lot more horsepower in these data centers to be able to handle the volume of
data that's coming.
>> So you guys obviously with the photonics and the liquid cooling, you guys have been
working on this in your labs for a long time, it's great R&D, but you need the connective
tissue because with 5G you're now talking about a ubiquitous RF cloud, powering autonomous
vehicles.
We're seeing the Brain Project here, ET Brain, the City Brain--
>> Yes.
>> Which is essentially IoT and big data being a big application that they're showcasing.
What's the connective tissue?
How does that work, from the data center, to the Edge?
What's Intel's position?
How do you see it?
And what's going to unfold in front of our eyes?
>> Yeah, so two things, so number one, I believe that the data center is boundary-less.
It's not based on four physical walls.
It's a connected link between the data center, and all the Edge devices that you called IoT.
In order to fulfill this, you have to have 5G technology.
We're invested in Silicon, in radio technologies, as well as in driving the 5G industry in consortia,
to be able to bring 5G solutions to market.
We think that 5G, as well as a tiered architecture between the Edge to the center, where you
do some processing at the Edge, the radio stations, some in intermediate data centers,
and then some in the back end Cloud data center, is what's going to be essential, and Intel
has significant investments, both in developing this distributed hierarchical architecture,
as well as in 5G.
>> That's a great point.
I want to just unpack that, and double-click on it a little bit, because you mentioned
data at the Edge, and you also said earlier, low latency.
Okay, a lot of people have been talking about, it costs you speed and time to move data around.
So there's no real one general architecturing, where you have to kind of decide the architecture
for the use case.
>> Yes.
>> So, the beauty is in the eye of the beholder, whoever has the workloads or the equipment.
>> Yes.
>> How do you look at that, because now you're thinking about, if I don't want to move data
around, maybe you shouldn't, maybe you want to move data around.
How does that fit with the Cloud of model, because we're seeing Cloud being a great use
case for IoT in one instance, and maybe not in another.
How do you think about that?
How should practitioners think about the data architecture?
>> Yeah, so our vision is that the Cloud changes from a centralized Cloud, to a distributed
Cloud, and is amorphoused between the Edge where the IoT devices are, and the backend,
and the way to think about it perhaps, is to say that storage as people have envisioned
it, as being centralized, that paradigm has to change, and storage has to become distributed,
such that data is available at different points in the network, and my vision is that you
don't want to move data around, you want to minimize data movement for most use cases,
and you want to have it pre-positioned on the 5G network, and you want to move the compute
to the data, that's more energy-efficient.
>> So I got to ask you, as someone who's doing the path-finding, which is the future path
for Intel, and innovation and architecture.
I was talking with some practitioners recently at another event, and trying to find someone,
because I don't speak Chinese very well.
But they asked me the same question.
It matters what's in my Cloud.
And what they mean by their Cloud, either on-premise private Cloud that they're putting
together, operating model of their business, now going Cloud-like.
But also as they pick their Cloud provider, they want to have multi-Cloud, and so what's
in their Cloud, and their Cloud provider's matters.
You guys are the inside of the Cloud across many spectrums, Intel.
>> Yes.
>> How should a customer think about that question?
What's in my Cloud?
Why should it matter, and it should matter.
What's your take on that, and what should they look for?
>> Yeah, so my take is that for years we've had the debate of whether it's public Cloud,
or private Cloud, or on-prem Cloud.
Our view is that the world is Hybrid, which is why we are big supporters of Alibaba, and
the Hybrid Cloud movement, and as such, if it's Hybrid, it sort of suggests that the
end state is that there'll be about an equal amount of applications that run on public
versus private, and so I think the number of applications have an affinity to move into
the public Cloud, like mail, and then there's other applications that you might care more
about the compliance and security that you would say have an affinity to being on-prem.
>> Also you mentioned that there's no walls, it's boundary-less in the data center.
Okay, there's no door, there's no mote, you can't put a firewall on that door, unlimited
access surface area for security.
Obviously security hacks are big.
We found out today that Israel had hacked, and notified the NSA.
Hacking is a huge problem.
Equifax is going to be another one.
How should customers protect themselves?
>> It's a very fair question John.
This is one of the side-effects of saying that the data center will be boundary-less.
We now have to have security technologies that can, we've effectively expanded the attacks
of security in a significant way, but I don't think the answer is to say we need to move
backwards and not adopt this boundary-less Cloud.
I think we want to adopt it, and we want to develop technologies.
So at Intel, we are developing multiple isolation technologies that allow different VM and container
tenants to be isolated from other tenants.
>> And this was your point earlier, making the device more intelligent, whether that's
more on-board memory, and more chips.
>> Yes.
>> That's what you were kind of referring to, is that right?
>> That's correct.
>> Okay great, so I want to get one kind of off-the-wall question, since I have you on
here.
It's just a brain trust here from Intel, which it's great to have him here.
Distributed computing has been around for awhile, we know all about that.
Network effects, distributed computing, the computer industry.
But now we're seeing a trend with decentralization.
Blockchain is one shining example.
Russia just banned cryptocurrency.
This poses a architectural challenge.
What's your thoughts on the decentralization, and distributed architectures that are emerging?
Opportunity is scary.
How should customers think about decentralization?
>> Well certainly there's a security challenge, as we just spoke, related to this.
But I think the computer industry has oscillated, depending on the era and the needs between
centralized and decentralized a number of times now.
And we're going through an era where decentralization makes sense, because we expect 30 to 50 billion
devices at the Edge, and so you can't handle that with a centralized model, primarily due
to three reasons, number one, just moving that volume of data would be very expensive
to do over the network.
Second there'll be a number of applications that are latency-sensitive.
And third, you might care about data federation, and crossing country boundaries in a number
of cases.
So I think for the use case that we have with IoT, we have to adopt decentralized and distributed.
>> So, if The Brain is processing and data, and you've got plenty of it at Intel with
more compute power, what's the central nervous system, the metadata?
>> Well, actually look at the central nervous system as the 5G distributed network that
enables the end-points, or the nerve endings if you will, to be connected to the spinal
cord.
>> Okay so a final question for you, I really appreciate you spending the time.
>> Sure, it's been a pleasure.
>> Intel's been a wave company in its generation, and obviously Moore's law, it's not well documented.
It seems that Moore's law is every year some journalist claims Moore's law is dead, and
that it never goes away, so we expect more and more innovation coming from Intel.
You guys have surfed many waves.
In your opinion, what waves are coming?
Because it feels like the waves are big now, but a lot of people think that there's bigger
waves coming.
That the big wave set is coming in.
What's the technology wave that you're looking at from a path-finding, innovation standpoint,
that customers should look for, maybe prepare for.
It could be further out coming in.
What's the big wave coming in, obviously AI was seeing these things.
What's your focus on that?
>> So, a number of them.
I think, you know distributed computing is not a solved problem yet.
But certainly it needs to be solved to be able to address these end-point challenges.
Another great example I think, is around visual computing.
So in the past, most of the type of data that people handled, was textual.
But that's moving to visual very rapidly, and there's so many examples.
You brought up the City Brain Project as an example.
But video and analyzing images, requires a different kind of art.
Different compression techniques.
If a human doesn't need to see it, you perhaps don't have to have as high a resolution, and
so there's a number of ships in the assumption space.
And so I think for me, visual computing is a great opportunity, as well as a wave, that's
coming at us.
>> And the software too.
So the final question, final, final question.
Alibaba here, are connecting the dots.
You can see where it's going.
How do you see the Cloud service provider opportunity, because obviously they're a Cloud
service provider on paper, but they're big, they're a Native Cloud now, like with the
big guys like Amazon, Google, Microsoft.
But we're seeing an emergence of new class of Cloud service provider.
Certainly our research is showing that what was a very thin neck in the power laws, now
expanding into a much bigger range, where VARs and value-edited software developers
are going to start doing their own Cloud-like solutions with the Native Clouds, because
they need horizontally scalable data infrastructure, connective tissue, and Edge devices from Intel,
but they're going to provide software expertise that's vertically specialized, whether it's
traffic, IoT, or oil and gas, or financial, Fintech.
The specialism of application developers combined with horizontally scalable Cloud, it seems
like a renaissance in the Cloud service provider market.
Do you see that as well, and how should the industry think about this potential renaissance?
>> So I think there's two possibilities.
One is for the vast majority of functions that people run in the public Cloud, I think
one possibility is that there's a consolidation amongst a few players.
But I think your point's a very good one.
That they are specialized services that companies are able to provide, where they're able to
carve out a niche, and become a Cloud provider for that particular set of functions, as well
as there's a second reason that motivates regional Cloud providers to succeed, again,
because of data federation requirements, as well as local proximal, proximity to the end-points.
I think these two phenomena are likely to drive the emergence of regional Clouds, as
well as specialized Clouds, like you described to perform certain functions.
>> And potentially a new kind of ecosystem development.
>> Yes.
>> And this is, then you guys are all about ecosystems, so is Alibaba.
>> That's right.
>> Dhiraj, thanks so much for coming on theCUBE, this is exclusive CUBE coverage with SiliconANGLE,
and Wikibon here in China with Intel's booth here.
Talking about AI, and the future of the data center and Cloud.
I'm John Furrier, thanks for watching.