00:00hi everyone welcome to the a6 & Z
00:03and today Michael and I are sitting down
00:05with our newest general partner Martine
00:07casado who is going to be covering all
00:09things infrastructure which sounds like
00:11a be really boring topic but it's
00:12actually a very interesting topic
00:13especially given his background which
00:15started off at Lawrence Livermore
00:17laboratories and then went into
00:18infrastructure after 9/11 after we
00:21briefly covered his background we talked
00:22about other interesting themes like the
00:25as a service if ocation of everything
00:27how open-source is actually more
00:29important for reasons that we don't
00:31quite realize and why selling to
00:33developers is so different so at that
00:35let's just get started
00:36Martine welcome happy to be here he is
00:45happy and he's gonna get happier I can
00:47tell you as this interview goes on or
00:49maybe the opposite as we start asking
00:51lots of questions my team you're on you
00:53eternal partner and you are gonna be
00:55covering everything infrastructure yeah
00:57also interested in so focused on to
01:00begin with infrastructure and in general
01:02Enterprise it's funny because we were
01:04just joking about this because
01:05infrastructure I think is the sexiest
01:07topic on the planet but I think the
01:09majority of the world and I'm rolling my
01:10eyes yeah it's boring being and yet it
01:15is really interesting has something
01:17changed in the evolution of computing
01:19that infrastructure is suddenly sexy
01:21even outside of just this room I mean or
01:23have you been hanging around with Peter
01:24Levine too much I actually think that
01:27the reason that infrastructure does get
01:30a lot of billing is because we don't see
01:31a lot of it right I mean like we focus
01:33on consumer as we focus on what we buy
01:34and what we can use day-to-day but
01:36anytime you shine a light on general
01:38infrastructure whether it's IT or not
01:40people find it super fascinating so this
01:41is gonna sound really silly but I'm
01:43gonna give you two reasons have you ever
01:44seen that TV show really big machines
01:48TV shows like these a really big machine
01:50mining equipment from you know South
01:53Africa it's like it's like like boy you
01:56put my daughter there and she thinks the
01:58most amazing thing ever because and this
02:00is just big piece of infrastructure and
02:02the reality is the most amazing
02:04engineering and technical marvels go
02:07into infrastructure we just don't see it
02:09I mean another example is I got a book
02:11on how to build a modern city it's
02:13infrastructure and if you look at kind
02:15of this shift that's happening in IT in
02:17Silicon Valley of course in the popular
02:20press you see it from kind of the
02:21consumer viewpoint cuz that's what we
02:22think day-to-day but if you want to see
02:24where the really magic happens and a lot
02:26of the the big technical innovations
02:28happen to get the infrastructure level
02:29so I do think it's very sexy natively
02:31did you did you always have a love
02:32affair with an infrastructure I mean
02:34let's talk a little bit about your
02:35background or did you did you arrive at
02:38it later yeah exactly I mean I doubt you
02:40woke up when you were like 6 years old
02:41and said I want to do infrastructure I
02:47mean sort of X I took a very crooked
02:50path in general to computer science and
02:52into infrastructure actually yeah I'm
02:55kind of a failed microbiologist and then
02:57a failed physicists well I heard that
02:58you dabbled in microbio physics
03:00astronomy I was thinking who dabbles in
03:02those things is that so funny yes
03:04actually I did um actually I started
03:05strongly I was at the local university
03:07Northern Arizona University I was taking
03:09classes and I was doing research there
03:10the interesting thing is I'd spend tons
03:12of time in this lab like you know and
03:14like basically we have to like check how
03:15bacteria were growing every two hours
03:17not spent all this time in the lab and
03:19in the corner was this computer and I
03:22actually didn't have a big background in
03:24computer I was more of a math guy and I
03:25just spent all of this time on the con
03:27the computer and so from then I said
03:29okay I was probably 20 at the time I
03:30decided to take computer science courses
03:32and because I had a physics background
03:34in mostly physics so I did two kind of
03:35physics computations so my first job out
03:38of undergraduate was at Lawrence
03:39Livermore we're toasting like massive
03:41physics simulations so at the time I was
03:43in Applied Computer Science in physics
03:46and not an infrastructure but I was
03:48there actually in the nuclear weapons
03:49program I was gonna say can you tell us
03:51what you were modeling or what what what
03:53physics you were going after so I was in
03:54I was in the weapons program so it's
03:55doing like like like large simulations
03:57of all sorts of things that relates to
04:00the weapons program right at the time
04:02and I was doing that through 2001 when
04:069/11 happened I was actually at
04:08Livermore when 9/11 happened and there
04:10like that moment the entire temperament
04:14of the the country change which is if
04:17you think about the weapons program its
04:18kind of an anachronism it's kind of like
04:20this holdover from like the Cold War
04:22that's exactly right I mean like listen
04:24that's a very different type of threat
04:27and what was now perceived as the modern
04:29threat and so the whole posture of the
04:31nation moved away from this kind of
04:32anachronism to the intelligence and so I
04:36was in this position where I had my
04:37clearances and I worked in a computation
04:39and environment and by the way the
04:40reason I ended up going to Stanford is
04:42I'm like listen I'm going into
04:43intelligence I don't understand
04:44everything so I started taking
04:45networking courses at Stanford no sense
04:48to me be you know modeling or simulating
04:51nuclear weapons programs to intelligence
04:55how does one lead to the next aside from
04:58your clearance here's my favorite thing
04:59about computer science and well always
05:01love computer science and this is gonna
05:02sound like an aside but I'll bring it
05:04back to what you said so here's what it
05:05is let's say that I write a program and
05:08it solves grand unified field theory and
05:11I've solved everything there's this all
05:12about physics physics would disappear as
05:13a discipline I would have solved it you
05:15know what I would say
05:16I'd say that's one more application out
05:17of go after biology and so in many ways
05:19I think that's so powerful that what
05:21computer science is so powerful at
05:23software's you can apply to so many
05:24different things and I've always been a
05:25guy about fundamentals I was really
05:27interested in the fundamentals about you
05:28of computer science there's two mindsets
05:30of how people approach this whole topic
05:32of computer science as Alan illiteracy
05:34one is that it's a language and you have
05:37to learn it just like English and and
05:38mathematics and anything else
05:40and another is that it is a tool for
05:41solving problems that can be applied in
05:43any particular way and I think that's
05:46interesting because that leads to people
05:47thinking about it more as a means to an
05:49end versus the end in and of itself it
05:52means like you can apply that mindset to
05:53anything in life yeah so I've always
05:54been exceptionally comply applied when
05:56it comes to computer science like I've
05:58never thought of it as actual science I
05:59mean we say computer science of us out
06:01of it's an engineering discipline and
06:02the goal has always been to solve a
06:04real-world problem so let's back up you
06:06sure you were working on intelligence
06:08you realize that wait I don't understand
06:11the fundamentals of networks and and or
06:13maybe there's a better way to do this
06:14the intelligence community's a lot about
06:16how people communicate what they're
06:17communicating the data understanding how
06:21our network infrastructures are how do
06:24you break into systems how to see use
06:25secure systems and so much of that is
06:27the network right systems are connected
06:29to the network and if you're going to
06:30break into something remotely you have
06:31to understand the network what was it
06:33and what made you realize that you you
06:34could head in a different direction yes
06:36so um I think the work in the last ten
06:38years like as I'm giving this podcast in
06:40the last 10 years of
06:41and trying to reimagine the the network
06:45with software and and all of that was
06:48rooted in this experience in the
06:49intelligence community yeah tell us more
06:51about that because I think it's actually
06:53hard to even go back to that yeah and
06:55you're right people take a lot of stuff
06:56for granted because it's different now
06:57and by the way most people don't even
06:59see it probably the first big AHA was
07:03the following which is if you're the
07:05government you have deep pockets and and
07:08you're going against some pretty
07:11intimidating adversaries nation-state so
07:13who've also have deep pockets
07:14normally market forces doesn't create
07:17things that you can use so let's say I
07:18want to build a real secure
07:19infrastructure well on the compute side
07:21you could buy computers and then you
07:23could program them for what you needed
07:25to for this very different environment
07:28this very different threat environment
07:29and there's there's there's ways to do
07:31that and so the intelligence community
07:32would do this they have modified
07:33operating systems for example SC Linux
07:35came out of this time but no it came to
07:38networking it was very different
07:39so in networking basically whatever you
07:43bought that's what you got and my bot
07:45you mean with computers you buy hardware
07:47but then there's a programming model so
07:50that you can write software on top of
07:51that so so you could for example buy a
07:53server get Linux and program Linux but
07:55when it came to networking you basically
07:56you'd buy the hardware it would come
07:58with software that was already written
08:00by the vendor and there was no model at
08:02all for you to evolve it you know one of
08:04my Jobs was to look you know to look
08:06around and and to determine whether
08:08different networks and environments were
08:10secure and how could you make them more
08:12secure many of the problems came down to
08:15the networking piece and to answer the
08:17question how do you make it more secure
08:18the answer is like well with the current
08:20technology to kind of can't
08:22so you started this this very simple
08:24thought in my head it's like you know on
08:26the on the compute side I can take it I
08:28can program it to do whatever I want but
08:31when it comes to networking I can't and
08:33this was the seed of what you know
08:34became Sdn and all the work that we did
08:35so Bastian you mean software-defined
08:37network sorry it's software to find
08:38everything and define that well at this
08:40point I think the terms become so
08:42diluted as to be meaningless so if you
08:44hear the term Sofra Defined Networking
08:45now it's now the the byproduct of a
08:48bunch of marketing departments that will
08:49apply it to everything right and so even
08:51I don't know what it means anymore but
08:53at the time it meant something very
08:54simple simple so when you know I did
08:57this work at Stanford for my PhD we just
09:00set the following a there's no
09:02programming model for networking it just
09:03doesn't exist and so there's kind of two
09:05things that you want to do to to get one
09:07one of them is you want to make the
09:09networking hardware sufficiently general
09:10purpose so that you have like an
09:12instruction set you have like the x86
09:14for networking so that you can program a
09:16single switch the second thing is
09:18because most networking problems don't
09:20deal with just one switch they deals
09:21with networks of switches they deal with
09:22networks of things is you want to have a
09:26discipline or a way of programming a
09:29collection of these things that you can
09:30show some stronger properties than you
09:32could before so we kind of said two
09:34things said one make an individual
09:36switch more general and programmable and
09:38two let's focus on a programming model
09:40that would go across them so you can do
09:42things like security things like
09:43mobility things like operational
09:44simplicity it reminds me of what
09:46happened earlier a couple a generation
09:49ago in the field of robotics where
09:51people used to hard-code everything into
09:54the actual robot and then they have this
09:56realization where you can actually make
09:59the hardware itself more generic to your
10:01point sort of using just as an analogy
10:03and then use a software to manipulate
10:05everything together I think there's a
10:07bunch of analogies that's a great one so
10:08so I think the classic analogy is back
10:11in the 70s you bought a mainframe the
10:12mainframe had everything in it you know
10:14and that's what you got and then the PC
10:17revolution came or you decoupled the
10:19operating system you could write your
10:20own operating some like Linna stood the
10:22robotics one is fantastic
10:23if you go 20 years ago if you wanted to
10:26make a robot like you have that I have a
10:28machine shop and today you can get a 3d
10:30printer and to get the components way
10:32back way and like to get a sensor would
10:33cost tons of money and today because of
10:35the advent of the iPhone you know I mean
10:37these things are pennies on the dollar
10:39so it's it's the ability to make your
10:42own components and then the components
10:44become very cost effective the exact
10:45same thing is happening in networking so
10:47before to get a router you know you pony
10:49up a hundred thousand bucks and like
10:51basically you couldn't modify it today I
10:53mean you can get 48 ports of 10-gig x'
10:56for $2000 and you can run your own
10:58software on its we're seeing the exact
11:00same type of renaissance happening
11:01generations of companies that were
11:02hugely valuable Cisco being the most
11:05obvious but you know companies that made
11:07switches that were you know
11:08just billions upon billions of dollars
11:10in sales yeah and that all went away I
11:12don't know if it's thanks to you but
11:13because the the switch became generic
11:15and that whole layer of gear became
11:17generic well I think we're seeing this
11:19playing out in real time I mean you know
11:21there's been some very active
11:23competition you know at the hardware
11:25level right now but we've seen like a
11:27big change in in you know the value of
11:30these physical switch vendors we'll talk
11:31about providing the way that they market
11:34and even the way that they sell because
11:37of this effort and so I don't think
11:38we've seen the end and I don't think we
11:40really understand the broad impact but I
11:42do think that we understand that it's
11:43massive it's changing the conversation
11:44with customers is changing buying
11:46patterns changing how we think of the
11:48technology and indeed it's changing how
11:49we create the technology and if you went
11:51back 10 years when this all started
11:53it's a different planet so where's the
11:54significance of the company that you
11:56co-founded with Nick McCune and Scot
11:59Schenker was a three and it was nice era
12:02and it was acquired by VMware which is
12:04where you just came from that's right
12:06what was the significance of this era
12:08so Sdn you can think of as a high-level
12:10architecture it's not a product it's not
12:12something that you create and give to
12:13somebody it's an architecture and in
12:14many ways it's a way to think about
12:16things right like like networking
12:18scientists by the way really love
12:26architecture I've defined a networking
12:31it's so obsessed about the ways of
12:34thinking of things right but you're
12:37saying is just an architecture why it's
12:39just like your programming isn't a thing
12:41you know programming is a way to think
12:43about creating things so Sdn is is is is
12:45yeah it's an architecture that it's a
12:46high-level thing so there was there was
12:48two kind of a hoss that we went through
12:50with nasarah so we knew that Sdn
12:52provided a nice way to think about
12:53things but we wanted to apply it to
12:55solve a real problem so people could
12:56actually do something with it that's
12:57right that's right to actually create a
12:59product have some real impact and look
13:00back and say you know the world is
13:01different because of this specific thing
13:02that we did and so so there were two
13:05trends that we took advantage of and
13:07they kind of just helped describe what
13:09we did so the first one is if you looked
13:11at mega datacenters the Google's the
13:13Amazons the Facebook's they were
13:14building networks very different than
13:17anybody else and so if you looked at
13:19what they did there they're building
13:20very very simple physical
13:22works and they're moving things that we
13:24typically put networks like like
13:25security and load balancing and a lot of
13:29the operations and management fault
13:32isolation instead of having that in the
13:33hardware they were moving it to the edge
13:35and doing it in software this is a
13:37massively big difference so now they can
13:38just by any simple switch that they want
13:40configure it very very simply and
13:41they've and then of the application they
13:43were evolving it to have all this
13:45functionality so that's to be an Sdn
13:47level shift so that was the first thing
13:49we noticed and these were by far by far
13:50the most efficient data centers on the
13:54planet run by you know I would consider
13:56some of the most technically savvy
13:57people and nobody sold them that
13:59architecture like they I mean they
14:00basically looked at the problem and said
14:01this is the best way to solve it and at
14:03the time and still to this day it was
14:05like their their secret like that's what
14:07differentiated them from everyone
14:08exactly yeah so I mean I considered
14:11Google and then and Amazon to be pretty
14:14much the leaders and you know this is
14:15ten years ago that they were doing this
14:16stuff and and and they'd really realized
14:19that you could just build very simple
14:20physical networks and put everything in
14:21in software at the edge so that was the
14:24first big vector that we're looking I'm
14:27like wow you know the world's gonna
14:28change once this gets out everywhere the
14:30second one was compute virtualization so
14:32at the time computer virtualization was
14:34really I mean it it was somewhat mature
14:37but it was still in this hyper growth
14:38phase I mean like you know VMware went
14:40public around this time and the
14:42interesting thing about compute
14:44virtualization is for every virtual
14:45machine there's a virtual port which is
14:48a network port and these are running on
14:50servers at the edge and I remember once
14:52looking and I'm like you know what
14:53within a couple of years VMware is gonna
14:55have more virtual ports than cisco has
14:57physical ports that's right because
15:01you're like listen these things are on
15:02the server in software the most rad
15:06datacenters on the planet have taken the
15:08application in software and change them
15:10to consume these things and the majority
15:12of workloads are gonna be on
15:13virtualization and they're gonna have
15:14software ports so what we thought to do
15:17at nasira's to take virtualization build
15:20out networking functionality in software
15:22so that any enterprise not just the
15:26googles and Amazons could take advantage
15:27of this architecture right the only
15:29reason that google's and Amazon's can do
15:31what they did is cuz they own the
15:32application they could rewrite the
15:33application but for City Bay
15:35or JPMorgan Chase it has tens of
15:37thousands of applications written by
15:39third parties written by Microsoft
15:41written by whoever we could provide that
15:44software layer that allows them to run
15:47very simple physical switches and move
15:48everything in software so it was a
15:49really alignment of two massive massive
15:51vectors we took about amazing shift and
15:54then you were acquired by VMware when we
15:57when we launched the company actually
15:59went public with the company not like
16:00públicas as in the stock market but we
16:03actually came out of stealth we had 18 T
16:07NT T eBay Rackspace and Fidelity's like
16:09five of the largest companies and it was
16:11your thesis true I mean did they get
16:13that immediately that hey the world's
16:15headed this way and we'll actually pay
16:16you to help us get there you know I I am
16:18still to this day so surprised by not
16:25only how quickly they got it but how
16:26fundamentally they understood it some of
16:28the people that I was working with then
16:30Toby Ford JC Martin Eric Karlin I think
16:36now looking in retrospect were such
16:38visionaries in understanding this change
16:40this huge risk oh yeah in some ways
16:43arguably even more so because it's a lot
16:45tougher for someone inside a captive
16:47system who already has a point of view
16:48and how to do things to embrace
16:50something new and different than for
16:52someone who's grading something they
16:55didn't build and when the rest of the
16:57world didn't get it because again this
16:59is a time nobody said Sdn right from a
17:01vendor no one had heard of I mean it
17:03didn't rhyme with IBM or or anything
17:05right yeah no right like the term Sdn
17:09like either hadn't been coined or like
17:11you know wasn't around very long and I
17:12mean these people really did see it
17:14understand that consume it and many of
17:16them I think most of them or all of them
17:18are still large customers and so we did
17:21that we started getting very good
17:22traction and then in 2012 we got
17:24acquired by an interesting little
17:26anecdote is that during because of that
17:29VMware was actually in Gartner's quad
17:32famous quadrant noted as a visionary in
17:35networking well so I mean there's many
17:38remarkable things about that number one
17:40this is a hardware quadrant it's the
17:42only pure-play software company ever
17:44how does some security play into all of
17:46this I mean I know we hear about
17:49very different way now than we ever did
17:50even two years ago yeah but networking
17:54at the security level are they the exact
17:55same thing I mean I almost feels that
17:56can you even separate those two topics
17:58anymore that's an exceptional
17:59exceptional question so so the simple
18:02one of the simple answers is is networks
18:06have to touch everything and security
18:07has to touch everything right and so
18:08just as you could put networking in the
18:11hypervisor to touch every single
18:13workload you could put security there
18:15and have security services and many
18:17security services operate at the network
18:18level which is basically who can or
18:20cannot talk to who traditionally
18:24security has been done on the perimeter
18:26of data centers right so the 80% of the
18:29security spent is at the perimeter of a
18:31data center even though a minority the
18:33traffic actually ever leaves the data
18:35center which means if you get beyond
18:36that like when I worked you know in the
18:38intelligence agencies if you get beyond
18:40the perimeter you have unfettered access
18:41to all the data and all the computes so
18:44I'm gonna tell one quick anecdote so so
18:46okay as the time of this recording I'm
18:4839 years old when I got turned 37 on my
18:51birthday my sister sent me an email and
18:54it was a it was a great email is like
18:55you know happy birthday brother I'm so
18:57happy you're my brother there's pictures
18:58of us when we're kids pictures of us and
19:05she said you know I'm very happy to have
19:08hung out with you last week we did hang
19:10out last week and you know you click on
19:11this link for my birthday it was awesome
19:13and like my first thought was it was not
19:15your thought my first thought was like
19:18my sister's never remembered my
19:32seriously like I've been waiting for
19:34this moment all my life yeah no but for
19:41the reason you know but then I'm like
19:42okay this is this a little this is a
19:45little strange even though there is
19:46pictures of us when we're kids or so far
19:48I mean like and so I looked through the
19:50Bale headers and of course it come from
19:51Russia but listen if my sister was not
19:54forgetful and I wasn't like a semi
20:00technical dude right like him either of
20:01those two things weren't we're
20:03true like I was wondering what your
20:04sister was doing in Russia the reason I
20:14say this is at the time that that email
20:16came in and I was reading it I was
20:18actually logged in behind the firewall
20:19in the data center I was like at work
20:21and so if I would have been infected
20:24that would have been infected behind the
20:25data center and like so many of the
20:28attacks that happened recently that
20:29would have been game over pretty much
20:30whoever would have gotten in would have
20:31gotten behind the data center so you
20:33know I just testified in front of
20:35Congress two weeks ago on exactly this
20:39problem which is you need to push
20:41security all the way into the data
20:42center they're no longer just securing
20:44the premium not just the perimeter and
20:46you need to do in the way that's that's
20:47you know reasonably secure from the
20:49computer science standpoint and you know
20:52this Sdn type approach I believe
20:54provides viable underpinnings for doing
20:56that isn't going deeper into the data
21:00center for security actually inviting
21:02more vulnerabilities it depends on how
21:05you do things so the first one is I
21:06wouldn't ever get rid of the perimeter
21:07the perimeter is great so the job of the
21:08perimeter is to keep every sociopath
21:11from the planet outside of your data
21:12center yeah but it may or may not be
21:13effective so you still want to keep the
21:15perimeter there right it's like um
21:17either/or that's exactly so so you want
21:19to add things to the data center also so
21:22so so it's purely additive in that sense
21:23but you don't want to add it at like an
21:26application level you don't want to like
21:27change the app and you do want to have
21:29application security but that isn't
21:31sufficient because once an application
21:32is attacked then that's like putting
21:33like an on/off switch on the outside of
21:35a house or something right that for an
21:37alarm system so once the application is
21:38attacked you could just turn that off so
21:40you want to have an infrastructure level
21:41that is outside of the hands of the
21:43attacker that's a different attack
21:45surface that provides ubiquitous
21:47security within the data center I mean
21:49that's the idea in some sort of medieval
21:51castle model you know there's cool walls
21:54but you just keep building walls or I
21:58love this now how did we like build
22:00cities in the 1300s when we built these
22:01really big walls and these moats and
22:03then basically hoped that nobody came
22:04over them right but then like you know
22:06in the modern day of like airplanes and
22:08you know whatever like those turned into
22:09imaginal lines if you know like you know
22:11World War 2 right you know you just kind
22:12of fly over them you parachute down so
22:14you basically need to put a sentry at
22:17by the way this is not this is not any
22:18company or product specific thing this
22:20is a general attitude a mindset right
22:22and a mindset yeah about security that's
22:25being adopted kind of kind of broadly
22:27right now and you see many many
22:29companies going down this approach but
22:30we do have to change security and to me
22:33the best way to articulate the objective
22:36is the following I believe the best
22:37security follow something called the
22:39principle of least privilege what is the
22:41principle the principle of least
22:45privilege is the following you can get
22:47exact access to exactly what you need to
22:50get the job done and nothing more so
22:52normally so let's take the day so you're
22:55talking about permissioning kinda like
22:56almost like permission levels I'm just
22:59saying that I will only give you access
23:00to exactly what you need to get the job
23:01done and nothing more so in the case of
23:02the data center if I'm a server
23:04I should really only have to talk to
23:06other servers that I need to talk to
23:09nothing else but the reality is is you
23:11know if I'm able to compromise a server
23:13I can talk to anything that I want and
23:15so we need to move infrastructure to
23:18this principle of least privilege so
23:21that for example if I do compromise the
23:23server I'm limited to only the things
23:24that that server can talk and so that's
23:26the for me the right mental mind frame
23:28from an infrastructure perspective to to
23:30build more secure systems but the
23:31reality is that because of things like
23:34virtualization and everything else
23:35applications are not neatly contained in
23:38like this linear way like what a single
23:40server set up for a single application
23:42that model has gone away so how do you
23:44then get from the principle of least
23:45privilege to match the reality of how
23:47applications are built this is a
23:49brilliant question like the way that I
23:52think it no seriously so the way that I
23:54think of it is the following which is
23:55the application has evolved into a
23:57network I really believe that so an
23:59application you'd be like I'm running
24:00pac-man but that's no longer the case
24:02now like think about like think about
24:04like one query to Google man like you're
24:07literally touching hundreds of compute
24:10you know databases balancers fire all
24:12these things are touched in order to get
24:14that one request done and as a result
24:18you need to evolve infrastructure from a
24:20point solution like a firewall to
24:22something that underlies every one of
24:23those components so have that have the
24:26security and infrastructure model adapt
24:28away to where how applications are
24:30actually built today
24:31trying to take on this thing that would
24:32be impossible to change which is not a
24:34design how applications are built yeah
24:36that's right so I think so the
24:37application has moved from a single
24:39instance to a network of things this
24:42kind of tightly coupled network of
24:43things we're even loosely coupled
24:45networks of things you want the
24:46infrastructure to have enforcement
24:47points that can wrap around all of those
24:50so you can treat the entire thing as a
24:51whole and secure the entire thing as a
24:53whole no matter how distribute it is
24:54I mean infrastructure is a service
24:57discipline right like we build sidewalks
24:59and and and buildings and cities yeah I
25:02mean that's that's what we do so so but
25:04we wanted to adapt to the populace of
25:05the populace adapts if I'm not
25:07stretching the analogy too much our
25:09populace is the application so the
25:12application has moved to this
25:13distributive thing and data centers now
25:15we need to evolve the infrastructure to
25:16have the same type of kind of
25:17flexibility and so forth to be able to
25:19map to that and though that is why
25:21you're seeing this massive renaissance I
25:23believe in infrastructures because we're
25:24where we add a dead run to keep pace
25:26with the application guys the whole I
25:27mean I think it is important to take a
25:28step back and I should talk about how
25:30computing has changed so much things
25:31that we take for granted I mean we have
25:32way more data than ever before it's
25:34real-time and faster than ever before um
25:37talk about what's coming next
25:39like what do you think is how things are
25:41changing like one thing that fascinates
25:42me is micro-services architecture that
25:45is of course the question and I actually
25:47think the major vectors aren't
25:50necessarily technical in the way we like
25:52to think so I think I think it's cool
25:53and listening I mean like you know like
25:55being a CTO for a long time and and and
25:57having an I love to think about like all
26:00the cool new stuff that's happening but
26:01if I look at what are the major shifts
26:03in the industry it's not super technical
26:05in as much as the following we're seeing
26:07this massive shift and then I asked
26:08myself what is this shift well it seems
26:10to me that we're seeing a couple of
26:12trends number one app developers are
26:16starting to consume infrastructure as
26:18pieces of software that's not really a
26:20technical thing as much as as a shift in
26:23responsibility away from something that
26:25used to be operated by one guy to
26:28becoming basically an object in a
26:30program created by another guy like
26:32Amazon for example or Google whomever
26:34yeah exactly so yeah so these guys are
26:35actually so the developers are starting
26:37to become those that like create
26:39infrastructure and they're becoming part
26:41of programs that's a massive shift to
26:44for example vendors write adventures are
26:46used to selling to IT people and now
26:48they sell to everyone now they have to
26:50sell everyone and now the buyer is is is
26:51different and and developers are very
26:54different than IT folks right developers
26:56love open-source they love as a service
26:58they've got a different aesthetic they
27:00got a different way of thinking about
27:01the world and so if you're used to like
27:03I mean the traditional IT sales model is
27:05like sales guy huge expense account
27:08briefcase and so we're seeing this
27:10massive shift you know towards the
27:12developer we're seeing and as part of
27:14that you know open source becomes very
27:17why does a source become so important
27:19because as part of their aesthetic right
27:20they like to maintain things I'd like to
27:22understand what they're developing so I
27:23think that's one of the big they get a
27:25huge shift another huge shift is very
27:27similar as a service shipping software's
27:30so hard because you you you give this
27:34offer to them and then you can all these
27:36versions out there and then you have to
27:37maintain then you have to update that
27:39and if there's a problem you have to go
27:40on site to find that doing it as a
27:42service it's so much easier because it's
27:43localized right which is if I like
27:45infrastructure services taking off so
27:47much so I mean like these different ways
27:50of delivering software these different
27:53ways you know the different buying
27:54centers I think this is kind of where a
27:55lot of this change is happening and so
27:59like I'm really you know like you know
28:01having been at VMware for for almost
28:03four years and being across thousands of
28:06customers I mean it's just so apparent
28:08that this shift is happening yeah I
28:10think actually I'm really glad you
28:12pointed that out because we have a
28:13tendency to get so caught up in the
28:14technological aspects of things that we
28:17forget the business model and
28:18organizational structural implications
28:20of things like software as a service how
28:23it reshapes companies buying processes
28:24sales everything around it and I also
28:27think it's really interesting you
28:28focused on something we're really
28:29interested in which is talking about
28:31selling to developers as a new class of
28:33sales what are some of the other things
28:36about selling to developers you've
28:37noticed well I mean it's just a it just
28:39everything about it is different like
28:41marketing is different right it's no
28:43longer about you know like a you know a
28:48marketing campaign where you go out
28:50there and you bring a bunch of customers
28:52in and you give them a class or so forth
28:53things like actually having open source
28:55out there that developers are going to
28:56use on their own whether
28:57home demo and test and buy as per the
29:00buying process yeah absolutely so so so
29:03open source becomes very viable
29:05marketing still like open source is
29:06marketing never nobody thought of it no
29:08but it totally is right I mean it's it's
29:10mean think about how many companies have
29:12been created out of successful open
29:14source product right yeah it's a trap
29:15before you buy it yeah yeah yeah I mean
29:20it basically creates your early customer
29:22base here or you know it does all of
29:24this sourcing for you but also gives you
29:25the credibility like a traditional IT
29:28sale briefcase walk-in talk to the IT
29:31buyer do some ROI calculations have a
29:34nice dinner a great bottle of wine you
29:36know you try and figure it out you get a
29:37deal done talking to the developer I
29:40mean they have to understand the
29:41technology right this is their
29:42livelihood to like integrate these in
29:44their things they have to maintain it
29:45all the time so generally if they don't
29:48know what you're selling a developer
29:49they haven't used it before they don't
29:51understand the implications just hard
29:52but if you wrap it around an existing
29:54open source project they'll have seen it
29:56already comfortable so fascinating
29:58because what you said about how many
30:00company what you guys are saying about
30:01how many companies have been built on
30:02open source what you're saying is
30:04actually more precise because most
30:05people have a tendency to conflate an
30:07open source project with oh it didn't
30:09lead to this company but what you're
30:10really saying is something very nuanced
30:12and different about open source as a
30:14vector to to the buying process it's so
30:17just such a difference and what's
30:20interesting about this is there's
30:21implications as to what type of company
30:23you build as a result I still believe
30:25the entire industry is trying to
30:27understand what it means from a business
30:28model perspective so you kind of solve
30:30your sourcing problem with open source
30:32you know you can you can source and be
30:35formal marketing something you can
30:36source and get early customers you get
30:37traction it solves the insertion problem
30:39well the typical insertion problem is
30:41very hard to get inserted like if they
30:43don't if you don't have an existing
30:44relationship if you're not IBM and you
30:46don't have IBM account control you're
30:47not Sisk you don't have cisco control
30:49actually getting inserted it's very very
30:51hard open source often solves the
30:53insertion problem they're already
30:54dealing with you know like the developer
30:56is already using it at home and using it
30:58in Amazon and using it anywhere else so
30:59so you're already there used it at their
31:01last job or whatever that's right
31:03you're already inserted you're already
31:04there it's a massive marketing thing but
31:07there's implications on the back end
31:09which is a how do you build a viable
31:10business around that
31:11Silicon Valley loves software companies
31:14because you get a reputable product to
31:15market with a really high margin and
31:17high multiples but if you start looking
31:19around a lot of the open source business
31:21balls it looks a little bit like a PSO
31:23company which is lower margins and
31:25sorry professional sales or professional
31:28services company it looks more like you
31:31know here's the software is for free but
31:33we're gonna give you professional
31:33services as part of that engagement
31:35that's a much more expensive thing to do
31:37right it's it's harder to scale from a
31:39company standpoint and so I believe the
31:42entire industry is going through this
31:43kind of this this this period of inquiry
31:46where we all realize that open-source is
31:48part of the buying process if you want
31:51to ship software what does that mean to
31:52the business model standpoint which i
31:54think is a fascinating question now as a
31:56service is really nice because you get
31:59like you look like a software company
32:00and you don't have to do a PSO model so
32:02maybe the answer is that most things
32:03will become as a service over time tell
32:05us a little bit about your
32:06transformation from PhD to CTO a lot of
32:09our founders go through that very
32:10process where they've been doing
32:12something they're really passionate
32:13about it whether it's through school or
32:15they just immerse themselves in it by
32:17just learning by doing and then to
32:19actually become the leader I think a
32:22major leader at the company like was
32:23that a big shift for you so you see the
32:25CTO it was straightforward it's becoming
32:28a business leader I'm a GM now right I'm
32:30really I'm a business manager for you
32:32know nearly a thousand people six
32:34hundred million dollar run right
32:35business and that's the transformation
32:36that's been more different
32:38and so I mean I think here I think I was
32:42actually basically forged in a volcano
32:44which is it sounds like the ring so in
32:55PhD which is just I mean it's just
32:57school right and you're just spending a
32:59tremendous amount of time thinking about
33:00one thing the problem is while you're
33:02doing that you get very refined in what
33:03you think about right you know and so
33:05you go to do a startup company and again
33:06this is the heady days of 2007 raising
33:08money was so easy we didn't really know
33:10like we had this idea but we didn't
33:11really know exactly what that looks like
33:13and then I just got you know hit in the
33:16face by a chainsaw when the market
33:19imploded so here I had 12 people I'd
33:21convinced him to join they had families
33:23they were having kids
33:24and it was my responsibility it was a
33:27really hard time I mean Sequoyah have
33:29released their rest in peace good times
33:31slide deck so many companies were going
33:34out of business like every week another
33:36one was going out of business
33:37unemployment shot up it was the worst
33:39housing market since the Great
33:40Depression and at that point in time I
33:43basically had a real basically hit with
33:46the realization that everything I
33:47learned the PhD was a distraction and I
33:49really mean that as a distraction which
33:51is pretty much every motion every motion
33:54of mine needed to be doing something
33:57incremental and doing something concrete
33:59to keep people fed which is exactly the
34:02opposite P so P PhD is about I'm gonna
34:04do a Hail Mary and honestly maybe you
34:08know three generations ahead of somebody
34:10who will care but like if someone cares
34:12now it's it's not novel enough and then
34:14I was stuck in a situation where I felt
34:15so deeply responsible for a set of
34:17people and I knew that every motion had
34:20to be something that was obvious low
34:22risk and was incremental and get his
34:27customers get us funding you know I mean
34:30it was just it's just such a different
34:32shift between let's do something just so
34:34abstract that you know finding people
34:36that care is hard to here's how we're
34:38gonna stay stay viable and so we spent
34:41two years at about 12 people if we
34:43didn't grow at all we stayed very
34:44focused and I mean so many times I was
34:46in the dáil the tailspin so many times
34:48in the void not knowing how we're gonna
34:50pull it out but through that experience
34:54I became exceptionally practical about
34:58it and then I think having the
34:59experience of the PhD and on the
35:02technical side just became something
35:04that I can use when I need to but I
35:06certainly don't use it away to guide my
35:07thinking especially not with with with
35:09with building companies and so now I
35:11find that a much much more practical
35:13much more focused on the business side
35:16just because of that this is a little
35:20bit like a homecoming and Ben and Mark
35:22invested in nice Sarah before and recent
35:25horse was even formed I guess and yeah
35:27and so what's it like coming home and
35:30and why like you moved out like why how
35:34did they make the decision to invest in
35:35you I mean you were a student weren't
35:38no no we did we just spun out we were
35:40still a research group you know what's
35:41interesting is the time that they
35:42invested in us was during the nuclear
35:43winter of 2008 if you guys remember and
35:46so like you know I kind of popped out of
35:48school in the heady days of 2007 the
35:51implosion happened in the end of that
35:52year and then I met mark and Ben after
35:54then in fact in fact I met mark for the
35:56first time in in NIEM I sat down with
35:59him and you know he had that little
36:00notepad that he still carries around
36:05like a new PhD kinda naive guy and I sat
36:08down and I basically said listen like um
36:11networking is gonna change I think I can
36:13show you all of these things if you if
36:15you live you know if you look behind the
36:16curtain at Google or if you look at
36:18what's happening with virtualization
36:19it's gonna happen for sure and he
36:21listened very very well it took very
36:24good notes but then he said listen you
36:25know we think that you know this is
36:26interesting so they invested this before
36:28the fund even existed honestly the the
36:30reason to come back to answer your
36:32question is is my excitement and the
36:34reason I'm going into VC is because you
36:37know in a transformation this size comes
36:39you can decide to participate along once
36:40a particular vector or across it and
36:42I've decided I want to have a broad
36:43participation I want to help I want to
36:45help fund the next transformation that's
36:47gonna happen and be deeply involved in
36:48that and then the way that I want to do
36:50this is the model that andreessen
36:52horowitz did just because I believe so
36:53fundamentally in the model having been a
36:55product of it margin we look we look
36:57forward to work with you more and
36:57talking more and thanks for joining the
37:00a 16z podcast for the first end of many
37:03times yes thank you welcome you guys