00:00hi everyone welcome to the a 6nz podcast
00:03I am sonal today's podcast episode is
00:05all about micro services and I've been
00:07super eager to focus only on this topic
00:09on the podcast since we mentioned it a
00:11lot in passing and I'm really excited
00:12because we finally get to do that our
00:14special guest for this topic is Adrian
00:16Cockcroft who helped lead Netflix's
00:18migration to a large scale highly
00:20available public cloud architecture a
00:22few years ago making Netflix one of the
00:24originators and early adopters of micro
00:27services and Adrian is widely credited
00:29for helping pioneer micro services at
00:31web-scale also joining the conversation
00:34or a 16z partners Martine casado and
00:36Frank Chen who will be moderating the
00:37discussion and in this episode we cover
00:39everything from what is micro services
00:42to the evolution of the architecture to
00:44how it changes the shape of
00:45organizations to operations to changing
00:48the role of CIOs and finally and this is
00:50actually what really excites me the most
00:51about this topic is what new
00:53opportunities come up when you have
00:54these extremely ephemeral systems that
00:56are you know just like like ghosts in
00:58the machine from containers to servers
00:59on demand to serverless and what's
01:01happening there and some really
01:03interesting trends on that edge the
01:04conversation begins however with the
01:06story of how Netflix got into micro
01:08services take us back to the days when
01:10Netflix had decided they were gonna move
01:12to Amazon and commit to a micro services
01:15architecture let's pick up the story
01:18there so what's it like inside
01:19we started off basically running away
01:22from a monolith we had over a hundred
01:25people every two weeks trying to get all
01:27the code they'd written in the last two
01:28weeks jammed into one codebase get it
01:31through QA and get that out into
01:33production and that was just getting
01:35more and more painful and we basically
01:38decided we had to break it into pieces
01:39he wanted it to be the work of one
01:42developer basically controlling what
01:44they had deployed independently of
01:47everybody else and at the same time we
01:49were looking at moving to cloud did you
01:51make both big moves at once in other
01:53words monolith the micro services and
01:55then private data center to Amazon
01:57everything together and sometimes you
02:00find incremental change a good theory
02:02but in practice you have to have a big
02:04enough stick to hit everybody with to
02:06make everything move at once and the big
02:09stick was we didn't have enough
02:10datacenter capacity to support streaming
02:12the DVD business in in the data center
02:16on a system that was growing at a
02:18respectable rate but the streaming
02:20business was exploding at a much much
02:22higher rate and because of that we knew
02:26we would have to either build lots of
02:27big data centers or get on to something
02:30else so the bet was okay we need to go
02:32on cloud then what's the right
02:33architecture for doing that what's the
02:35right organization for doing that the
02:37developer group is getting bigger and
02:39getting less productive and we wanted to
02:41unlock the innovation so we were
02:43simultaneously trying to get better
02:45developer productivity better time to
02:48value of just one of the key things
02:49we're trying to optimize for generally
02:51and then there was a whole bunch of
02:53other cloud transitions bundled in as
02:55you went from the monolithic application
02:57to microservices what did that entail
02:58what's that mean what is a
02:59micro-services architecture well
03:01originally I called it fine-grained SOA
03:04service-oriented architecture and
03:06there's a lot some people get negative
03:07reactions to SOA because they were out
03:10there trying to do it 10 15 years ago
03:12that's right so with all the same ideas
03:14over and over again with new dressing
03:15yeah there's a question like why now and
03:17why didn't it work then and if you look
03:19at it what we were doing was on
03:21relatively slow CPUs compared to what we
03:24have today on relatively slow networks
03:26we were processing big fat lumps of XML
03:28and passing it around and we were really
03:31only able to break the application into
03:33a few large chunks because the overhead
03:36of all of the message parsing was too
03:38high if you come to today the you know
03:41you can break it into you know maybe
03:43100th of a size and 100 times as many
03:45chunks because the overhead of the
03:47communication is now very low we've got
03:49binary protocols we're not trying to
03:51sort of make everything conform to the
03:53big soap XML messaging schemes so it
03:56became possible to build a fine-grain
03:57SOA architecture and that ended up being
04:00called micro services but I think Fred
04:02George was the first to use the word but
04:03that got written up by Martin Fowler and
04:06then everyone said ok and we'll go with
04:07that yeah so Big Bang moves this was a
04:10bet the company set of Technology
04:13looking back at it what are some of the
04:15lessons learned I think one of the ways
04:17to approach this is to basically create
04:20kind of a pathfinder or a pioneer team
04:23there was a lot of controversy in sites
04:26and he thought this was stupid yeah a
04:27few of us thought we could make it work
04:29and other people a bit more gung-ho so
04:31we got the people that thought they
04:32could make it work into a room and had a
04:35one-day project where we all built a
04:37thing in the cloud to see if it would
04:39work based built out of the kind of
04:41technologies we'd need to use to build
04:43this that team then sort of knocked down
04:46a bunch of the strawman arguments that
04:47everyone else was holding up against us
04:49a lot of the time it is just straw man
04:51argument but you have to actually go and
04:53build something to actually feel that
04:55find out what are the real arguments and
04:56then you discover things you didn't even
04:58know which are hard you know you run
04:59into the real blockers as opposed to the
05:01imaginary ones so I think the trick is
05:03to get a small team go very deep
05:05discover what you can and run a whole
05:09bunch of these little projects where
05:11you're trying to learn as much as
05:12possible with the smallest possible
05:14input you had this cultural aha which is
05:17let's get the people who are gun ho
05:20about this and let's let them go deep
05:21knock down the strong man arguments sort
05:23of zoom up to the 30,000 foot view and
05:25sort of describe the organization at
05:27Netflix sort of before and after what it
05:29looked like before and after from a
05:31skill set point of view from an
05:32organizational design point of view this
05:34is actually one of the big things that
05:36makes a difference some organizations
05:37are set up already to do micro service
05:40based architectures and others have to
05:42go through a reorg right Netflix it ever
05:45emerged naturally out of the way we were
05:47structured at the time we were already
05:49structured as small cells that owned
05:50things a lot of responsibilities each
05:53team had a very clear idea of what it
05:55was building and how it related to other
05:57teams but it was assembled as a monolith
06:00at the end of the day so breaking it
06:02apart was a fairly natural thing for us
06:04to do what you see with traditional
06:06enterprise siloed organisations as
06:08they're actually having to do a real gun
06:10setup teams that own that are
06:12responsible for services and that's it's
06:14somewhat unnatural for the way they're
06:16currently set up but I'm seeing an
06:18increasing number of people go through
06:20that transition and sometimes you can
06:22see it as a replacing project-based work
06:26with product based work so every team
06:28becomes basically a product team for
06:30their micro service and you have the
06:31product management aspects and the
06:33operational aspects within that team and
06:36did you find that the people who were
06:38used to working on the monolith could
06:39retrained or did you have to have a new
06:41crew come in in the culture and
06:43Netflix's is interesting most of us had
06:45been around before a lot of us had
06:47worked on SOA we you know the
06:49gray-haired people that have been
06:50there's a few people that work at Xerox
06:52PARC in the 1980s and you could go and
06:54have arguments with them about
06:55object-oriented programming we had some
06:57young younger people but there was a lot
06:58of very experienced people taking all
07:00the stuff they'd learned and
07:01synthesizing it together it's very
07:03collaborative experience and we came up
07:06with things that made sense based on
07:08this this series of transitions we were
07:10going through the other transition was
07:12from a single centralized database we
07:14had this enormous Oracle machine with a
07:16really complicated schema to a
07:18distributed no sequel database of in the
07:22end based on lots of different Cassandra
07:24clusters and that that was the third
07:26transition and that was probably the
07:27hardest transition was getting all of
07:29the sequel code and transactional stuff
07:31out of the system it's actually breaking
07:32apart the databases probably the hardest
07:35thing to do and then splitting chunks of
07:37code offers is also difficult if you're
07:39trying to pick apart our monolith and it
07:42turns out if you don't break apart your
07:44and you just creates or services to talk
07:46to it you've actually created what's
07:48called a distributed monolith which has
07:50all the same fragility of the monolith
07:52and you can't update things
07:53independently because you're tied by the
07:54database you can't just take the Oracle
07:57database and break it up into little
07:59pieces you have to think about it
08:01differently now the same thing is true
08:02for the rest of the architecture as you
08:04migrate to microservices yeah so I think
08:07what excites me about micros services in
08:09general it moves all of infrastructure
08:13up to an application layer so if you
08:16think about what you normally do an
08:18infrastructure you've got these basic
08:19abstractions I compute and network and
08:21storage which are pretty low level and
08:23there's semantic free right you don't
08:25have structured data one of the the huge
08:27advantages is going up to a micro
08:29service architectures you can do
08:30infrastructure insertion things like for
08:32example security things for like you
08:34know even debugging basic operations and
08:37management and you can do it in a way
08:38that has the deep context and semantics
08:41of the application the point here is
08:44that not not only are you going away
08:46from the monolith which is really
08:47important and I think it's great but
08:49also like you've got more semantics than
08:52before I mean this is actually
08:53meaningful stuff when you're dealing
08:54with not IP headers for example not
08:56blocks but actual like structured data
08:59and I think that we can actually
09:00reimagine a lot of these tools in ways
09:02that we've never thought of them before
09:03because we've never had the ability to
09:04have this type of semantics in this in
09:06these tool chains we're seeing this
09:07burgeoning area of micro services where
09:09you almost have like a function per
09:11company coming up and now I believe that
09:12all of the old stuff that we had in the
09:14internet whether it's naming your
09:15service discovery or routing or whatever
09:18we've got an opportunity to bring this
09:19up in kind of a much deeper richer
09:21richer level which is which is really
09:23cool right so we were going to the
09:25market place where the bazaar away from
09:27the Cathedral which is any any
09:29individual function can be provided by
09:31either an internal or external provider
09:33it could be a cloud service but then the
09:36challenge is now it's up to every
09:37organization to coordinate right and so
09:40what are some lessons that you guys have
09:42learned along the way of picking
09:43best-of-breed and then making sure they
09:45work with each other getting the version
09:47control to work when you've got a
09:48monolithic app everything is in there if
09:50it gets broken into you have all access
09:53its connection to the database lets it
09:55basically say anything to the database
09:57when you break things into micro
09:59services you've got the ability to have
10:01some parts of your system be low low
10:03security risk and other parts be high
10:05security of risk you can innovate really
10:08really quickly in areas of sort of
10:10personalization and user experience and
10:12then you have maybe have a much more
10:14tightly controlled thing for say the
10:17signup flow and that way you're storing
10:19personal information so the great news
10:20is you have a lot more agility the price
10:23that you pay is you're doing a lot more
10:25coordination with the monolith it's easy
10:28you put all your eggs in one basket and
10:30then from a security point of view for
10:32instance you basically just pile a bunch
10:33of appliances in front of it easy right
10:35because it was a monolith he knew
10:37exactly where it was now that the
10:39perimeter is distributed across many
10:41machines you have to be a lot more
10:42mindful of where the attack surface has
10:45gone and which security service you need
10:47to put in front of that part of the
10:49micro-services architecture so that you
10:52can not have the privilege escalation of
10:54because there is a little bit of PCI
10:56compliance needed in one tiny corner of
10:59this monolith the entire monolith is now
11:01subject to PCI compliance and Sox
11:03compliance and all these things and by
11:06two pieces you can have most of your app
11:08be extremely agile and very innovative
11:11and then have the bits that need to be
11:13safe be extremely safe and then if you
11:15look at the attack surface you're
11:17basically keeping a very tight control
11:19over what can do what and if you connect
11:22into the databases you've got very
11:23single purpose connections into the
11:25database that are doing one thing you
11:27can start to control at the access level
11:29there as well what used to be policy
11:32controlled by the operations people what
11:34they felt was a safe sandbox for the
11:36developers is now really being driven
11:38from the other end down so this idea of
11:40developer driven infrastructure is
11:41something that is turning things around
11:43and a lot of what I'm seeing is a big
11:45banks and people like that they have
11:47their existing policy frameworks and
11:49rules and they're trying to apply it in
11:51the new world and it looks the same so
11:53they're happy because they're compliant
11:54but they don't actually have the real
11:56policy separation that they used that
11:58they think they have because it's all
12:00totally reprogrammable and it's like you
12:02have the illusion that you're still
12:04conforming to the to the policy a lot of
12:06these things were ops controlled so the
12:08ops would control the data center and
12:10then the network's in the data center
12:12and now it's all developer defined and
12:16software constructs which are controlled
12:18by your cloud API s but if you're
12:20updating it ten times a day there isn't
12:22time to have ten meetings a day with
12:24operations to do the handoff so what
12:26we've been seeing is that people just
12:28running it themselves the only person
12:30that knows it's the exact state of the
12:32system is the developer that just
12:33updated it that sounds scary until you
12:35realize that each of them is controlling
12:37a very small piece of the system and the
12:39aggregate behavior of the system turns
12:40out to be really robust and reliable
12:42partly because if you put a developer on
12:45call they write really reliable code and
12:46they don't release code on Friday
12:48afternoons and because they want to
12:50quite weekend and you know they learn a
12:51bunch of practices about having a what
12:54it's like to be on call and help not to
12:55break things so we went from an
12:57in-person Change review board
12:59infrequently right into the vet the
13:01changes to continuous change and Pig
13:04let's coordinate over slack pretty much
13:07yeah you have to tell people what you're
13:08doing but you don't have to typically
13:10ask for permission and go and have
13:12meetings and things like that
13:14this is part of unlocking the innovation
13:16and people are most interested in this
13:17are large teams of people
13:19trying to build complex products
13:21typically enterprises and they are
13:24worried about getting disrupted by the
13:27latest Bay Area startup or whatever
13:29there's an existential threat here that
13:31if you're doing quarterly releases and
13:33your competitors doing daily releases
13:35and continuous delivery you're gonna
13:36fall so far behind in the user
13:38experience that you're just gonna suffer
13:40right so that's the big driver that is
13:43making people say well how do you get
13:45there there's a whole bunch of things
13:47tied together you're bringing in cloud
13:48DevOps is a whole other area and micro
13:52services as an architecture all these
13:54things tie together and some cultural
13:56change as well in the organization of
13:58the company the companies that are doing
13:59well at that are really starting to
14:01accelerate off into the distance
14:02it's also worth teasing apart to trends
14:05and and and one of these trends is you
14:08know a single company instead of
14:09building a monolithic product wants to
14:11build a micro services product and
14:12they've gates all the efficiencies of
14:14doing that as far as the development
14:15process and the OEM process and
14:16everything else but there's kind of a
14:18broader industry trend where companies
14:21products are basically micro services
14:23right there's companies out there that
14:25like basically the only way to access
14:26the product is through a fairly narrow
14:28API I mean you know there's so many of
14:30these now that there are other startups
14:32that will just basically stitch them
14:34together and you could build full
14:35applications without writing much code
14:37so I think that in addition to a single
14:39company getting a lot of advantages I
14:41think the entire industry is gonna get a
14:42lot of advantages and see a lot of
14:44innovation as a result yeah if you had
14:46said five years ago that there would be
14:48multiple independent public companies
14:50that all they do was offer an API you
14:53would have been left out of the room
14:54right right and now look at us Twilio
14:56and stripe and on and on I like to do
14:58the mental exercise of kind of where
15:00this is all going and I I still loved
15:02Chris Dickson's quote of you know every
15:04UNIX command becomes a company as I crab
15:06becomes Google or whatever like I think
15:07you know we may be having an analog here
15:09which is every function becomes a
15:10company right more more granular than a
15:13command-line tool every single function
15:15or logical function becomes an
15:17independent company and I do think there
15:19are implications on things like
15:20ownership and dependability and and
15:23stuff like that that we haven't grappled
15:24yet as an industry but it's a very
15:25exciting direction yeah you're able to
15:27build something now that pulls in things
15:30from api's and pulls in some
15:34containers and you just have your little
15:36piece of code in the middle that
15:37stitches it together and build a
15:38completely new service from that so it's
15:40just much easier to get things built
15:42it's more efficient for the big
15:43companies but that's it has democratized
15:45all the way down to pretty much anybody
15:48with a laptop can Co build build
15:50something interesting and if you go back
15:52five or ten years that you're doing
15:53things that would be just totally
15:54impossible to try and get together at
15:56that point there's much more room for
15:58innovation it also makes it harder to
16:00compete in some ways because now it's
16:02hard to build you know a billion-dollar
16:04software company on top of these things
16:06because they keep changing underneath
16:08you and they're cheap to build so you've
16:11got lots of disruption coming and it's
16:13actually you know github an open source
16:16is another big player in here that's
16:18just bringing making it much lower cost
16:21to get things done so what you're seeing
16:23now is Twitter and Facebook and Netflix
16:25and Google and LinkedIn producing the
16:28stuff you actually want to use which has
16:29already been tested at volume and then
16:31it's actually much harder to build a
16:33proprietary software company because
16:35you're competing with these big
16:37end-users and you've got this thing
16:38you've just built and it's flaky and
16:40doesn't quite work right we've talked
16:41about this it seems like closed-source
16:43shippable software is is on its way out
16:45or dead and there's a number of reasons
16:47for this one of them is just the
16:48enterprise buyer likes open source
16:49software but another one is it's a real
16:51burden on the company to ship software
16:53right I mean especially if that software
16:56is a distributed system right I mean
16:57like you don't have skilled operators
16:59often every environment is different
17:01right so you've got these heterogeneous
17:02deployment environments you end up with
17:03this thing like the mother of all cache
17:05consistency problems we've got a bunch
17:06of versions out there the product we've
17:08got to maintain bunch of versions it's a
17:10human matrix from hell right now the
17:12cubicles of version x the flavors of
17:14unix x whatever Windows versions you're
17:17supporting right or QA manager yeah
17:20that's right and then distributed
17:21systems generally like I mean like a
17:22real trick if you're running your own
17:24operation is you have skilled
17:26administrators that know how to manage a
17:27cluster and then like there are very
17:29very few companies and I think maybe one
17:31that's actually managed to ship a
17:33distributed system that was manageable
17:35with a non skilled operator it's a very
17:36very difficult problem and so a great
17:39thing about if you offer something as
17:42service is like okay you don't have any
17:44of these problems and so like basically
17:47operation budget is way lower it's much
17:50easier to start a company now but at the
17:52same time there are questions about like
17:53okay so like what are the size of these
17:55companies are gonna end up being like I
17:57mean like how big is the market for a
17:59single function I think it's still to be
18:00seen like how big these companies are
18:02gonna are gonna become yeah big
18:04challenge from an investor's point of
18:05view which is if the essential argument
18:06is there will be no more cathedrals it's
18:09all bazaars from here on out it's a
18:11little harder to make money truck and
18:14that's because it's gonna get yeah yeah
18:16yeah so put yourself in the shoes of the
18:18enterprise CIO the pace of change is
18:20accelerating right the ink just dried on
18:23her team getting VMware certified and
18:25now we're on to containers and then
18:27people are talking about server lists
18:28and functions at us as a service with
18:30sort of lambda architecture so talk a
18:32little bit about what's coming and then
18:34the ability of an average organization
18:37to sort of absorb these changes
18:38containers came along really over the
18:40last two years and one of the fastest
18:42takeovers of enterprise computing we've
18:44ever seen it's quite quite remarkable
18:46how quickly they were able to colonize
18:48the enterprise space it solve the real
18:50problem what role did containers play in
18:52moving away from the monoliths to the
18:54micro services architecture what happens
18:56with the containers all that stuff is
18:57packaged into a bundle which has all the
18:59right versions of everything inside it
19:01and you can download it and run it
19:03it also abstracts you away from the
19:05particular version of what you're
19:07running on and there's now containers
19:08for Windows as well but originally this
19:10was a Linux based concept you have the
19:12same container format if you want to run
19:14in house or on a public cloud it doesn't
19:17really matter that container can run on
19:19VMware or KVM on OpenStack or on Amazon
19:23or Google or Azure or wherever right
19:26you've just abstract youjizz up up one
19:27level it gives you that kind of
19:29if you think about machines used to sit
19:32at the same IP address for years people
19:34would know a machine they would actually
19:35know the IP address off by heart if they
19:37wanted to do something to it and then
19:40you had VMs came along and other VMs are
19:43more transient and they know this thing
19:44would come and go maybe in new order of
19:46weeks or something but by weekly update
19:48of your VM and then with containers it's
19:51perfectly reasonable half container that
19:53runs for less than a minute you can
19:54create an entire test environment it set
19:56it up run it run your tests you know
19:59automatically test it strip the thing
20:01and the the size of the things have got
20:04much smaller if you just take it to its
20:05logical conclusion we basically fire up
20:08effectively a container to run a single
20:10request and have a chicken kicks her out
20:12sit around for about half a second and
20:14then have it go away again and that's
20:16really what the underlying technology
20:18behind AWS lamdaur it's a server
20:20on-demand that just isn't there most of
20:22the time and this is the bleeding edge
20:24right now we have to figure out how to
20:26extract these sort of ghostly flickering
20:29images that are sort of coming into
20:31existence for short periods of time how
20:33do you track what's going on you end up
20:35doing figuring out how to do end-to-end
20:36tracing as the only way you can monitor
20:38things rather than being a special case
20:40like it is now so there's a bunch of
20:42interesting problems here but what's
20:44really been happening is just this trend
20:46to more and more ephemerality and these
20:49extremely ephemeral systems and then the
20:51charging used to charge by three years
20:53worth of machine and then it became
20:55while you can rent a VM by the hour and
20:57then containers you know that's light
20:59weight lighter weight and now you're
21:01paying by the hundred milliseconds right
21:03it's perfectly reasonable around for
21:04half a second which means that the
21:05set-up time to create that half second
21:09worth of machine needs to be radically
21:11less than half a second and the time
21:13taken to Bill it bill for it needs to be
21:15less than half a second if you remember
21:17the story of SMS the SMS record for is
21:20140 characters the billing record is
21:22much bigger than that it's like a
21:24kilobyte so if you actually take the
21:27telco and rip out all of the billing
21:29stuff for their SMS things you know it's
21:3210th you know it's it would cost a tenth
21:34of the amount to run if they didn't
21:35billed for it right so you get this
21:37effect that the overhead of doing the
21:39thing is actually vastly more than the
21:41thing you're trying to do so that so
21:42there's actually it's a really
21:43interesting challenge is how to create
21:45monitoring and billing and scheduling
21:48systems that work so quickly that you
21:51can afford to build things in tiny
21:52increments our portfolio company 21 is
21:55sort of right in the thick of this right
21:56which is how do you stand up and ad-hoc
21:58agreement between an API and an API and
22:00like have the billing all work and you
22:02know Bitcoin might play a part in that
22:04so also to your question going back to
22:06the CIO I mean it seems to me in general
22:08with disruptive technologies is like the
22:10disruption happens first and then all
22:11the day to ops happens second I mean
22:14that is and I think in this case name of
22:16the disruption is around delaminating
22:18the app and breaking it apart I do think
22:20that CIO should not despair enough steam
22:22should not despair because what happens
22:24very quickly in the vacuum being left
22:26from kind of the you know this this
22:28sprint on these new technologies is
22:30whole you know ecosystems and whole
22:33industries arise around them to provide
22:34visibility to provide security to
22:36provide ops and we're seeing that now
22:37and so I mean I think that it's quite
22:40possible to decouple the disruption
22:42which is this velocity around
22:44development and then you know the basic
22:46operations and that tooling was
22:48definitely going to happen as well and
22:49understanding that ecosystem an
22:51understanding the players I was very
22:52important if you want to stay on top of
22:53this kind of big change leaning forward
22:56into the change assuming the tooling
22:57will meet you halfway exactly and then
22:59you get the benefit the big benefit from
23:01the CIOs point of view in my opinion is
23:03that you don't have this loop where the
23:05business user asked for something it
23:07took you 15 months to build it only to
23:10discover that's not what the business
23:11user really wanted because the
23:12requirements are poorly specified in
23:14these days right no problem I've got a
23:16change for you we'll put it live this
23:18afternoon right so the rapid
23:20experimentation that happens in start-up
23:22land can now migrate into the big
23:24organizations and you don't have to get
23:26your requirements perfectly specified at
23:28the beginning of a waterfall process
23:29anymore let's run the experiments it's
23:31actually even better than that what the
23:33CIOs are providing now is a set of API
23:35is for the built for the development
23:37team that is part of the business to
23:39automatically provision whatever they
23:41want with certain policy constraints
23:43around it for what they can and can't do
23:45but fundamentally you're providing api's
23:47operations has moved from being a ticket
23:49driven organization to be an API they
23:52are now no longer a call center that is
23:55a very profound move and I'm seeing a
23:57lot of these CIOs buying into that they
24:00want to be part of the product they want
24:02to be how do you support the business
24:03and you provide API so that they can
24:06just get business done at a rate that
24:08you're not slowing them down we're
24:09actually seeing the creation of a new
24:11buying center in the industry I've her
24:13to call platform engineering I've heard
24:15it called the DevOps you know whatever
24:16this is like budget allocated it's
24:17actually viewed as a profit Center is
24:19product aligned but its court
24:20infrastructure and operations and these
24:22are very technical buyers so it's not
24:23the traditional enterprise go to market
24:26this is also moving across into
24:28trees we've seen of sea media and
24:30entertainment and to some extent retail
24:32were early movers mostly because of the
24:34threat of Amazon themselves crossing
24:36retailers to step up to reengineering
24:39we're now seeing FinTech fighting their
24:41Wall Street is really paying attention
24:43to some people are way down the road
24:45some people are just starting
24:46manufacturing that whole industry is
24:48just starting to think about this
24:50there's definitely a sort of industry by
24:52industry sort of domino effect as people
24:54are figuring this out so we're a decade
24:56on or so into this revolution right many
24:58strands what excites you now for me
25:01what's really exciting about this I've
25:03said this before is if we just have the
25:05ability to reimagine all of
25:06infrastructure you can now reimagine
25:09tooling and reimagine security and
25:11reimagine operations and management we
25:13get to reimagine it with more semantics
25:16and context than we've ever had you know
25:18so what does it mean to have a firewall
25:19in a world for everything as micro
25:21services what does it mean to have
25:22operation management and debugging
25:24things that were traditional boxes that
25:26were stuck on perimeters now also become
25:29functions and actually managing your
25:30infrastructure is almost like looking at
25:33a debugger a context debugger it's like
25:35you have a symbol table with you it's
25:37like this whole thing is in the one
25:38large IDE and you can do that for your
25:40operations I think it's gonna push the
25:42state-of-the-art on how we even think
25:44about offs in an entirely new areas I'm
25:46really excited about that change I think
25:48that the whole serverless area is the
25:50bleeding edge right now the monitoring
25:52tools industry is right now being
25:54disrupted pretty heavily by serverless
25:55there's only one or two tools that have
25:57really come into existence in the last
25:59year or two that have a effectively a
26:01way of processing stuff that this is
26:03this ephemeral and dynamic so some
26:05interesting products coming out it's
26:08just a better way of living if you're a
26:09developer and you're working in the
26:11waterfall siloed organization it's kind
26:15for a lot of people indeed and and when
26:17you get ownership of a product on
26:19distributed teams you get each
26:21distributed team their own product
26:23ownership and they get to define the
26:25interface and manage it and run it yeah
26:28you might be on call but you're a much
26:29more control of your destiny and it's
26:31much more rewarding and it's more
26:32productive and the ability to get more
26:34stuff done as a developer is just
26:36rewarding anyway right it's a better
26:38it's a better way of working for people
26:40well that's great well thank you
26:41dream Thank You Martine we'll see a lot
26:43more unfold as the architecture shifts