00:00hello this Peter from andreessen
00:02horowitz and I'm it's my honor to be
00:06joined with by a Ben who's the CEO of
00:10digitalocean Ben how you doing you're
00:14doing really well thanks Peter for
00:16good yeah great so one of the things man
00:21that I've I've spoken quite a bit about
00:23recently in sort of the data center of
00:27the future and the future of enterprise
00:30technology is the notion of open source
00:34economics finally working and
00:36specifically kind of contrasting open
00:39source being used being used as a
00:42service like what digital ocean does
00:45versus let's say the red hat model of
00:48you know on-prem free software and then
00:51charging for for you know the most
00:54charging for support maintenance as the
00:58economic model and my belief is that
01:00there's really a very beneficial model
01:03by having open source as a service open
01:05and proprietary source as a service
01:08again as a way to really drive economic
01:12value out of open source and I'm just
01:14sort of curious I mean you guys are
01:15you've built a company on this so I'm
01:18curious on your thoughts and how you
01:19guys think about that yeah definitely so
01:24we we built our cloud on top of kvms
01:28virtualization technology which in and
01:31of itself is an open source project and
01:33we deliver a ton more value than just
01:37the the virtualization engine so we
01:40created an entire interface and API to
01:44interact with the infrastructure that
01:46we're providing and we've also put you
01:48know thousands of servers into a data
01:51center now all of this is powered by by
01:55KVM and we've turned it into a
01:58infrastructure as a service provider
02:01without you know the work that was put
02:04into the open source project we'd never
02:06be able to build our company so I think
02:09this is a big widespread trend that
02:13that were seeing in in the open-source
02:17community first of all the projects
02:20themselves are becoming much more
02:21relevant to what people are arguing
02:24today but at the same time there's also
02:26a proliferation of new projects that are
02:29coming out every every month and every
02:32year so database-as-a-service
02:35essentially has has grown significantly
02:37over the last decade with a whole bunch
02:40of no sequel projects that are they're
02:44just you know getting their their start
02:46and what's interesting it is as an
02:49individual user it will take you a very
02:53long time to figure out how to leverage
02:55that project you have to learn the ins
02:58and outs there's a ton of best practice
03:00that goes along with it
03:02and one thing that the open-source
03:03community hasn't really done a great job
03:06of just yet is giving you a primer and
03:09getting you started on best practice
03:13implementation from day one and I think
03:16that's where this open source as a
03:18service provider really can step in and
03:22hopefully they've taken the time out to
03:25protect the technology on their end
03:26which is based on that open source
03:28project but it it's prepackaged for an
03:32end users need so now let's say I do
03:35have a need for a clustered scale-out no
03:38sequel database I can just go to one of
03:41these open source provider and and
03:43consume that service without taking all
03:45the time to build to build it on my own
03:48no this is definitely a prevailing trend
03:52that we're seeing more and more yeah so
03:54it's interesting and and you know maybe
03:57to contrast what like it it may not seem
04:02obvious to folks that ma it's not even
04:07completely obvious to me that open
04:10source is out there I can readily pick
04:12up open source pieces and I can put
04:15those pieces in my on-prem installation
04:18so like if I'm thinking about this y-you
04:23know what's the benefit of as a service
04:26you know just accompany themselves
04:29I'll say cobbling together open-source
04:32components they're readily available
04:33free to use maybe you know are their
04:36pros and cons to each of those and how
04:38might people think about adopting you
04:41know open-source as a service versus
04:43putting open-source on Prem yeah that's
04:46a great question a customer that I was
04:49just speaking to recently is is taking
04:52this new philosophy that they're
04:54building inside of their business and
04:56what they're looking for is the
04:58opportunity to outsource all of these
05:02open-source projects to a specialized
05:05provider and their basic premises we
05:10don't want to be responsible for the
05:14operational challenges that come with
05:16scaling and running that technology so
05:20if I consume it at the service I have a
05:25very fixed rate under which I pop it and
05:29it's going to cost me you know X cents
05:32or dollars per time period or per per
05:37volume or whatever the case might be and
05:40and and it really is a fundamental
05:45decision that a company needs to make do
05:48you outsource these pieces to
05:50specialized providers and hope that they
05:55provide a better level of service and
05:58you can develop internally you have
06:00actually someone next that you can choke
06:02especially because you're paying the
06:04money for that service yeah or do you
06:06build it in-house and with this specific
06:09customer they've already built a lot of
06:12this as a as a first pass on premise you
06:17know taking that open source and
06:19deploying it to their own infrastructure
06:22and then realizing that there's so much
06:24complexity that goes into managing each
06:27one of being discrete pieces and more
06:30importantly if you just take the open
06:34source project you don't have any
06:36commercial support that you can rely on
06:40system breaks it's up to your
06:42operational people to fix it and you
06:46know that was the decision point that
06:49they used to say in our particular case
06:51we would much rather outsource this to
06:55we believe that our total cost of
06:57ownership actually winds up being less
06:59and more importantly if something breaks
07:02we know that we're not on the hook so I
07:05think the example that he used was uh in
07:09this particular piece leveraging AWS is
07:12art yes product which is basically my
07:15sequel as as a service mm-hmm and he
07:19basically said look at the database is
07:21broken there's nothing that we need to
07:23do because we know that the service
07:25provider will go ahead and and fix the
07:28problem for us once the issue is
07:30resolved business is back online and
07:33operational whereas the opposite
07:36approach would be he needs to get in
07:38touch with his operations people and and
07:41make sure that they're responsible for
07:42fixing the problems so it's a
07:45fundamental shift in responsibility and
07:48and call us and does it make sense to
07:51outsource it to a specialized provider
07:53or deal with the operational challenges
07:56of running it in-house yeah it's really
07:59interesting I in addition to open source
08:03being offered as a service I've also
08:06spoken about and written about this
08:10whole notion of data center as a service
08:13where you know you just spoke about
08:15these data bases being offered as a
08:18service and if you think about it every
08:21part of the old on-prem ecosystem is now
08:26being made available as a service
08:28whether it's storage security
08:32infrastructure right all databases to
08:35your point are all now being made as a
08:37service and I'm just curious like you
08:39guys are in this business and you're
08:41you're absolutely you know driving that
08:44trend why what what is the catalyst that
08:48has made this happen now versus at some
08:52point in the past and why are we seeing
08:54all these sort of infrastructure
08:57components now being offered in the
09:00cloud and you know what's driving that
09:02trend there are a few things that are
09:07coming together at the same time so we
09:09have virtualization that really came out
09:13and in the mid-2000s and became
09:15mainstream I think that really sparked a
09:20lot of change in in the in the data
09:23center industry ok virtualization kind
09:26of became mainstream so what like what
09:30did that do to catalyze this on-prem to
09:33off friend movement because I think
09:34you're onto something and I'd love for
09:38yeah well part of the issue became that
09:41you have this substantially larger
09:44infrastructure to manage because now a
09:47single physical machine could run a
09:50hundred virtual machines and so
09:53organizations or businesses needed to
09:56become much better at managing all of
09:59these systems of their deployment and
10:01what happened is the open source
10:03community caught up to this need and
10:06they built a whole bunch of tools that
10:08allow that provisioning process to work
10:11a lot smoother so now you can actually
10:13provision to your machine on the fly you
10:16can maintain them inside of some type of
10:19inventory management system and you have
10:20full control and visibility over what's
10:22happening so virtualization created the
10:25opportunity for more servers to exist
10:27and there are physically in the world
10:29and then the open source tools came
10:31along to allow you to manage that effect
10:34effectively and I think those are pretty
10:36big catalysts to drive the beginning for
10:40you know this utility type of model
10:42where people can come in and because the
10:45part of the problem is in the old world
10:48essentially you would typically
10:50provision one dedicated server per
10:53customer and you wouldn't always have a
10:56need for the full resources that the
10:58Machine provides so there was a mismatch
11:00between what the service provider sells
11:04and then what the customer needs and so
11:08and being able to deliver a much smaller
11:10slice you've now matched supply and and
11:14demand and that allowed people to really
11:17consume these you know services on an
11:20as-needed basis because the tools and
11:23the technology enabled provisioning of
11:25really thin slices in an automated
11:27fashion and everything kind of just you
11:30know supportive support itself whereas
11:32in the old model you have people that
11:34would get a phone call and being told
11:37hey we need you to go ahead and rack and
11:39stack a server with this many processors
11:42this much memory and these disks that
11:45doesn't really allow this automated
11:47approach to work interesting so this
11:49fluidity of of sort of the backend
11:54system as enabled by virtual machines
11:56really created this this operational
12:00model that that that worked very well
12:03for multiple customers getting maximum
12:06use out of the out of the system in the
12:09backend and that's probably that you
12:11know that's what you're saying is real
12:13catalyst that's yeah
12:15yeah it's also interesting it's also
12:16interesting to know that in this
12:18particular case both extremes are
12:21serviced really well by this model in
12:23terms that a really small customer could
12:26now get started and enjoy the benefits
12:29of a world class enterprise
12:32infrastructure you know for for a few
12:35cents an hour and at the same time a
12:38large scale enterprise can really scale
12:41out their operation without incurring
12:43all the topics you know requirements
12:46that go into building out a huge data
12:49center and more importantly as business
12:51dynamics change and the workload changes
12:54they can quickly adjust you know in a
12:56matter of seconds rather than being
12:59stuck with hundreds of machines that
13:01they no longer have a use for so it
13:03created a lot of efficiencies from you
13:06know from the smaller cluster all the
13:07way into there into the enterprise yeah
13:10yeah let's maybe shift gears a little
13:12bit and talk about kind of the needs of
13:17the developer I mean your business is
13:21you know kind of anchored around a large
13:24large number of developers using your
13:27cloud to you know do their work
13:30why wouldn't developers just go to a you
13:34know sort of standard public cloud well
13:38I have a developer oriented cloud what
13:40are the differences and you know what
13:43happens over the course of time do the
13:46developers continue to use a developer's
13:48specific cloud or move to something more
13:50generic I'm just sort of curious how you
13:53think about different and then can we
13:55envision different clouds for different
13:59you know maybe developers what you guys
14:00are doing but we have you know a cloud
14:03for rocket scientists right and you know
14:07I don't you know exactly so like how do
14:13you think about that and I'm sort of
14:15curious how you see the world evolving
14:17and you know kind of the importance of
14:19having a developer oriented cloud versus
14:21a generic cloud and how you know how you
14:24guys think about that yeah oh prior
14:28products at digitalocean we feel that
14:30there was just the generic cloud and the
14:34analogy that we like to use is the you
14:38know today's cloud providers are really
14:40focused on the technology and they
14:44present that to their users in the most
14:47direct manner possible and what that
14:51allows for is kind of what we spoke
14:54about previously this really walled
14:56state allows infrastructure to develop
14:58it gives you all the knobs and controls
15:01that you would need to take you over
15:03obviously there is a bunch of
15:04operational work that's at the customer
15:07the user needs to need to do in order to
15:10take advantage of this public generic
15:14cloud but there's no limit to what they
15:17can accomplish the challenge that comes
15:20up is and as you build larger and larger
15:24scaling systems the complexity increases
15:28it's not even exponentially to deliver
15:32so what we recognize is that there was
15:35opportunity for us to focus on on people
15:39rather than on technology and because at
15:42the end of the day everyone is using the
15:44same virtualization engines there are
15:46two or three popular ones and it's kind
15:49of like you know silicon used to build
15:51professors no one has a substantially
15:53faster CPU than anyone else you know
15:56today and the same thing happens in
15:59virtualization everyone's virtualization
16:01engine is it's fairly similar in its
16:03capabilities and in its performance and
16:06so it boils down to what what are what
16:10are the core values that you're focusing
16:12on and as I mentioned most cloud
16:15providers today are trying to focus on
16:17technology and they think by introducing
16:19the next feature introducing the next
16:21service they can win a substantial part
16:24of the market and that is fairly that is
16:27a fairly accurate statement what they
16:29don't realize is that there's an
16:31opportunity to focus on on the people
16:34and who actually uses cloud and hosted
16:39well developers right they are one of
16:43the fastest growing segments of users
16:45online there are 20 million developers
16:47worldwide today and that grows that a
16:50couple million people per year and what
16:53we realized is that the actual human
16:56experience of trying to wrangle that
16:59infrastructure is is very painful and
17:02that the problem that digitalocean tries
17:05to solve is by presenting a simple
17:09intuitive easy-to-use interface to bring
17:13that infrastructure under control we use
17:16the same virtualization engines as the
17:18other clouds but where we spend a lot of
17:21our time is on designing the interface
17:24so that people can go ahead and
17:26provision these systems rather than it
17:28taking large enterprises and large teams
17:31of ops people and software engineers to
17:34figure out well how do I build the auto
17:36scaling group and how do I manage these
17:39thousands of servers across multiple
17:42what digitalocean is setting out to do
17:44is providing the easiest way for people
17:48having to hire even more people who want
17:52to run their operations in the cloud
17:54against that backdrop
17:56could you imagine other vertically
17:58oriented public clouds for other you
18:03know other folks out there you know
18:05could you imagine something you know
18:07beyond developers let's say not for
18:09digital ocean but in the industry in
18:11general does the public cloud vertical
18:14eyes around around people as you're
18:15saying like you guys have done or is
18:18there only one which is developers and
18:20then there's everyone else I think there
18:23is opportunity for other verticals to
18:26exist and depending on how far you want
18:28to stretch the cloud analogy you could
18:32say that company such as Squarespace or
18:36our Vic's you know these companies that
18:38provide build your website in the mod
18:45you know they cater to a specific
18:46segment of the market as well but
18:48they're certainly not an infrastructure
18:50provider they're more of a cloud website
18:54it's you know if I mean you've got term
18:58and and maybe even if we go back to
19:02infrastructure itself I think that there
19:05is an opportunity for for another focus
19:08to to exist but it has to be so
19:11fundamental in pristine suite a real
19:15connection with the users and to stand
19:18out right now if we say that current
19:20cloud from the generic providers are
19:22focused on technology and we all use
19:24that and we focus on people and
19:27obviously all customers are people at
19:29the end of the day even if it's a large
19:31enterprise like like Nike there's still
19:33someone that's making that decision
19:35finding another bucket
19:38that's as fundamental as those two I
19:41think it's possible but there aren't
19:44many of them and you know it takes a lot
19:47of thought and a lot of focus and what
19:50winds up happening quite often is that
19:52these buckets actually exist many times
19:55over but they're so small that in the
19:58grand scheme of things we we are not
20:01these companies so is there a third
20:05large bucket that a infrastructure
20:08provider could focus on I'd like to
20:12thing that there's always an opportunity
20:13to innovate to figure out the next step
20:16you know when we were building
20:18in 2011 this is the dilemma that we face
20:22how do we how do we build a company that
20:26stands out in in the sea of all the
20:29other clouds that already exist and it
20:31took us many months to be able to figure
20:34out our position around around
20:36simplicity and focusing on on users that
20:40are developers so I think if someone
20:42else took the time out to uh really
20:45digest the market and understand what's
20:47happening there's certainly an
20:48opportunity for her new entrance the the
20:51challenge is that quite often we find
20:53companies reacting to existing
20:56competition or or to market conditions
20:58rather than being able peers through
21:01that you know fog and and and truly
21:04create something new yeah let me um just
21:07sort of in rapid-fire here some
21:09competing things you know we talked
21:11about internally Linux containers are
21:14virtually or or virtualization what do
21:17you think in the future your waiters
21:24you know Linux containers started
21:26started the whole trend there's this a
21:29bunch of talks nowadays about the
21:31hypervisor is dead no I don't think
21:33we're there just yet containers have a
21:36very poor security model and that's
21:39really what's holding us back but
21:40there's certainly more lightweight the
21:42real difference between the two is the
21:45container has to take on the properties
21:48of its parent hypervisor so you're very
21:52restricted you can't today we're on a
21:54Windows container inside of a Linux
21:56server right and that's what
21:58virtualization allows you to do so it
22:00kind of depends on the use case for
22:01general public computing
22:04I think the container is the future we
22:06just need to fix the security model okay
22:08flash versus spinning disk well flash I
22:12think that's a no-brainer okay
22:15so disk goes away completely I think
22:19this is great for archival and and
22:23backup media because of its cheap cost
22:25and and fairly good performance and
22:28reliability especially in low io
22:31environments so I must say you know
22:34second year long long-term storage so
22:37desk becomes the new tape yeah
22:39okay Intel her arm Intel I wouldn't bet
22:44against them okay fair enough there you
22:48well good so with that band thanks a lot
22:53we're gonna you know call it a wrap and
22:56I'll have you on at some point again I
22:59love to see how all that materializes in
23:01the future so we'll talk with you
