( ESNUG 585 Item 6 ) ---------------------------------------------- [03/27/19]
Subject: Anirudh and Sawicki on why CDNS and MENT did the Cloud this year
DAC'18 Troublemakers Panel in San Francisco
Cooley: Anirudh.
Anirudh: Hey John.
Cooley: You just announced at this DAC that you're going whole hog into
the cloud. Synopsys did that in 2000 and in 2011 with something
called their DesignSphere.com site. How do we know you (CDNS)
are not making the same mistake?
Anirudh: First of all, the infrastructures have evolved a lot over the last
several years. But you have to look at this cloud thing in a more
holistic way. For example, I visited a pretty big customer and I
asked them how they are using their LSF farm. And I think about
60% of the farm usage was verification, doing logic simulation.
Then 20% or so was other verification, like timing verification
or circuit simulation, things like that. And the remaining 20%
was implementation and other things.
So that's not necessarily how they would spend their EDA money
but if you look at the usage of the farm, 80% were badge-oriented
verification simulation tasks. And they have to maintain these
huge farms and you have to update them every year. So, there is
a desire either for complete outsourcing or for peak utilization,
especially for things like verification and simulation to have a
cloud offering.
But I think that's just one part of the picture. Multiple things
have to come together. So one thing is a good infrastructure,
and I think Dean [Drako] talked about some pieces of that
infrastructure.
We have carefully worked with some of the leading providers like
Amazon and Microsoft and Google over the last couple of years to
make sure that the infrastructure is solid.
But the second piece, which is probably the most critical, is
the ecosystem and security. If I look at the past attempts, I
think most of them get stuck on this issue.
Because people have to trust that they can use the cloud. If you
look at HR systems or finance systems or even Salesforce, people
have established that trust, but when it comes to design data or
hardcore R&D engineering - that trust is still to be established.
Therefore, I think the ecosystem which is the second piece
becomes critical, and that's why we work closely with ARM and
TSMC. And you know at TSMC, both Cliff Hou and Suk Lee, took
the initiative. They ran several security audits on our system
and I think TSMC is quoted in our press release today.
So, I think the ecosystem and security is a very, very critical
piece. And I think that is what has prevented adoption in more
engineering functions of the cloud.
Cooley: Because they engineers are terrified of their stuff getting
stolen. I get it.
Anirudh: Yes. And then the 3rd piece is the EDA tools have to be equipped
to run in the cloud and we have made a lot of investments in
terms of making them massively parallel, so we can put the EDA
tools on the cloud. Also Cadence has a good business model with
the EDA card, maybe from Joe's [Costello] days...
Costello: Ancient history... (laughter)
Anirudh: So I think if you put all... You know but the EDA card is a very
good business model that Cadence has.
So, if you look at the infrastructure piece (which is important),
the security and ecosystem piece (which is I think is critical)
and the EDA tools and the business model -- you put it all
together -- and then we [Cadence] can finally offer a credible
cloud solution. And we'll see how the customers accept it.
Cooley: So, side question. Doesn't this mean your CDNS sales force;
your sales guys are going to go away, like travel agents went
away when Orbitz and Travelocity came about?
If I can download my EDA tool stuff why do I need the sales guy?
Anirudh: Well, we have to see how the use models are. ...
Cooley: Well you're doing a standard use model. You're doing per week,
per month, per quarter, per year...
Anirudh: Yeah, but it depends on the kinds of EDA customers. I think some
customers -- especially the top customers -- we just have to see
how the usage develops. Even in our current CDNS sales force
there are different use models. We have a distribution model,
where we work with distributors. We have a direct model. So,
we will have a cloud model to go with that.
So, it's a comprehensive sales strategy.
Sawicki: You know John, I think one of the things that happens in this one
is people say "cloud" and it becomes a magic word for them.
So, there was the first version of cloud [SNPS DesignSphere.com]
which you talked about where people who were attempting to build
essentially a Salesforce.com of EDA.
It was a walled garden. Going into that [SNPS] thing, people are
only going to access that thing. You get all your EDA tools from
them [SNPS]. It comes in with all the [SNPS-only] business model
aspects and the like.
And that is just weird with EDA ... mostly because of what Dean
[Drako] talked about which is every single one of the EDA
customers out here has this incredibly intricate [mixed EDA
tool] flow that they care about deeply -- because they think
it's their competitive advantage.
This cloud stuff that's mostly being talked about *now* is cloud
as a scalable hardware platform. I can't speak for it because I
only saw the CDNS press release, but it's a very sensible thing
that Cadence announced. Which is you have the EDA tool licenses.
You can now host them in the cloud as easily as you can host them
inside of your LSF world. And you can run your software out there.
And it's not a closed garden -- if you want to use other [non-CDNS]
EDA tools in conjunction with that you can put that in place.
And that's the direction that most of us have been going for a
while. We [at MENT] run stuff... we have customers running our
Mentor tools in the cloud today. My peak usage for doing test is
actually in the cloud, because we have 10,000 CPUs and some days
we need 20,000 CPUs.
Cooley: This is your own inside Mentor...
Sawicki: Yeah. Both internal, our our usage. And we have customers
that are on there as well. The aspect of that as a scalable
platform because of the investments that are being made to put
that in place -- it only makes sense. But it's not cloud as in
it's Salesforce.com...
Drako: Joe [Sawicki] is absolutely pointing out an important thing that
you gotta kind of look at as a basis. We started as SaaS -
software as a service - which is a Salesforce.com. And then
Amazon effectively introduced infrastructure as a service, and
that's just basically computers and object storage as a service.
And the computers and object store -- well the object store
doesn't work that great -- but the compute EC2 stuff works much
better for EDA, than the SaaS model. And so now Cloud is a lot
more viable.
Cooley: Okay. Joe... Sawicki... because there's two Joes. Why did
you launch Veloce in the Cloud (see ESNUG 583 #4) but nothing
else in the cloud this DAC?
Anirudh went whole hog. You guys went half-hog.
Sawicki: I guess there's an aspect of it... If I'd had read their press
release first maybe we would have done something else.
(laughter)
I don't know... I kind of struggle to kind of care about the
question. I can tell you why we did the thing with Veloce. It's
because it's a great big large piece of capital equipment that
for many customers, the scale-out of being able to do that inside
their own data centers is difficult.
This enables us [MENT] to go after a whole different set of
customers by being able to give them that on the cloud.
Costello: I think all of this we're talking about is right. It's obvious,
everything is going to be in the cloud 5 years from now.
Everything. I mean fundamentally, that is going to happen...
---- ---- ---- ---- ---- ---- ----
Related Articles:
Anirudh and Sawicki on why CDNS and MENT did the Cloud this year
Costello dissecting Montana, Rocketick, Palladium, Zebu, Veloce
28nm vs. 7nm, AMS, Virtuoso, CDNS Innovus vs. SNPS IC Compiler 2
Costello on EDA ossification, cloud, and RedHawk vs. GreenHawk
Join
Index
Next->Item
|
|
