2865: Navigating the Multi-Cloud Maze with CockroachDB

[00:00:00] Is your organisation ready to navigate the complex challenges of data management in a

[00:00:07] multi-cloud environment?

[00:00:08] And in a world where business continuity and resilience are so much more than just buzzwords,

[00:00:16] enterprises are increasingly turning to solutions that not only promise that uptime, that high

[00:00:21] availability, but also delivery in the most demanding conditions.

[00:00:25] And by that I mean when you've got 300,000 people on your website on Black Friday.

[00:00:31] That one day of the year where you make the vast majority of your profit.

[00:00:35] So today we're going to delve into the heart of this conversation with Spencer Kimball,

[00:00:40] CEO of Cockroach Labs, the visionary behind CockroachDB.

[00:00:46] And if that sounds familiar, that's because we spoke to him six years ago.

[00:00:49] So if you've got some long-time listeners listening and you might remember Spencer.

[00:00:55] But with database technology, not just surviving in the multi-cloud ecosystem but thriving, I

[00:01:01] think it's offering businesses unparalleled resilience, scalability and the freedom from

[00:01:07] vendor lock-in.

[00:01:09] That is the exciting stuff that I want to explore with Spencer today.

[00:01:13] Spencer will talk about how Cockroach Labs is helping to redefine what it means to

[00:01:17] be truly cloud native and why major companies from Netflix, Bolts and Comcast are betting

[00:01:23] big on this technology.

[00:01:26] So with his insights into the evolution of the database market, the strategic importance

[00:01:30] of data locality and the future of cloud data management, this episode is going to be a

[00:01:35] deep dive into the backbone technologies that ultimately power today's digital enterprises.

[00:01:41] But before we get today's guests on, it's time for a quick shout out to the sponsors

[00:01:45] of Tech Talks daily and in today's digital age where data breaches are all too common,

[00:01:51] and the common sense of information has never been more critical, right?

[00:01:54] Well enter KiteWorks, a pinnacle of managed file transfer security or MFT security.

[00:02:01] So with its FedRAMP moderate authorization, a prestigious certification that they've held

[00:02:07] since 2017 by the Department of Defense, KiteWorks stand as a bastion of security

[00:02:13] and a sea of uncertainty.

[00:02:14] So stepping to the future of secure file transfer with KiteWorks by visiting kiteworks.com

[00:02:20] today to see why it's held as the most secure MFT on the market.

[00:02:24] So once again, kiteworks.com and now let's get today's guest on.

[00:02:29] So buckle up and hold on tight because no matter where you are in the world right now,

[00:02:33] I'm going to beam your ears all the way to Stateside where Spencer's waiting to

[00:02:37] return to the podcast.

[00:02:39] So a massive warm welcome back to the show, Spencer.

[00:02:43] Oh, that spoke, I think something like six years ago.

[00:02:45] So can you just remind the listeners with a little of who you are and what you do?

[00:02:50] Absolutely. It's a pleasure to be back.

[00:02:52] So I'm one of the co-founders and the CEO of a company called Cockroach Labs.

[00:02:57] What we do is we make a database that's called CockroachDB.

[00:03:01] And as the name suggests, at least hopefully it suggests to people,

[00:03:05] it's a very resilient database.

[00:03:08] So it's difficult to kill and it can survive outages that would typically

[00:03:12] take down a more legacy traditional database architecture.

[00:03:16] And in the world of cloud, it's never been more important to build your business

[00:03:22] in such a way that when the cloud suffer a failure, which they do,

[00:03:26] and these are big headline items when it happens,

[00:03:29] there's an opportunity for you to have complete business continuity.

[00:03:33] And this is sort of a secret sauce in the database, but what the database does

[00:03:37] is it's replicating your data across different facilities,

[00:03:41] like physical facilities, different physical geographies,

[00:03:43] even across continents to give you and across cloud vendors

[00:03:46] to give you an unparalleled ability to keep your business running

[00:03:51] even when in the natural course of things, there are big outages.

[00:03:56] And I think the last time we spoke back in 2018,

[00:04:00] you shared the tech startup story of 3x Googlers and Cockroach Labs

[00:04:05] and everything that you were setting out to achieve.

[00:04:07] And fast forward to 2024, Cockroach Labs has now been identified

[00:04:11] as a key player disrupting the traditional database management market.

[00:04:15] So can you just tell me a bit more about the core innovations,

[00:04:18] the strategies that have helped position Cockroach DB as this preferred solution

[00:04:23] for so many huge companies and household names from Netflix to Boats

[00:04:28] and Comcast and so many more?

[00:04:31] Yeah, I mean those are just the companies whose logo we can talk publicly about.

[00:04:35] We have got a long list of the world's really big,

[00:04:40] important companies that are relying on us,

[00:04:44] key amongst all differentiators for this incredible resilience story.

[00:04:50] And when you think about the world's biggest companies

[00:04:54] that are offering use cases, whether it's, for example,

[00:04:58] retail banking or trading systems or logistics

[00:05:01] or there's some really big tech companies in there

[00:05:06] that if they went down, obviously it would impact everyone's lives

[00:05:10] in a, I guess just in the ordinary course of getting things done

[00:05:15] that need to get done in the dead.

[00:05:17] So result is that number one differentiator and many of these systems

[00:05:21] have run as well as they could have run in the past several decades

[00:05:27] on things like mainframes, which is a very venerable legacy tech stack

[00:05:33] produced by IBM and they're still heavily in use

[00:05:36] and it's a really excellent piece of sort of architecture

[00:05:40] and that within a data center provides almost unparalleled business continuity.

[00:05:46] The problem of course is that and the opportunity is that

[00:05:50] things go wrong on a geographic basis

[00:05:52] like a storm will take out data centers in Texas

[00:05:55] and then the question is does your business maintain uptime,

[00:06:00] availability, good customer interactions for the potentially millions,

[00:06:04] tens of millions or hundreds of millions of customers out there

[00:06:06] that are relying on you.

[00:06:08] And with the cloud in particular but also with the technology

[00:06:12] that cockroaches are bringing even ability to tie disparate geographies together

[00:06:17] to replicate across regions and so forth.

[00:06:20] So every resilience story is key among our differentiators

[00:06:23] but also in a similar fashion to how we can replicate data across geographies

[00:06:28] and really have this distributed nature to the system.

[00:06:31] We can actually distribute all of the scale that's necessary

[00:06:34] for increasingly data intensive applications.

[00:06:37] So many of our customers have, I mean,

[00:06:41] some of our customers I think with well into the hundreds of millions of customers

[00:06:48] it's not just those customers that are creating traffic on the system

[00:06:52] and demand for the data.

[00:06:54] It's also increasingly virtual representations of those customers desire.

[00:06:59] So think about AI agents and the like

[00:07:02] and all of these things in order to work in increasingly complex ways.

[00:07:07] There's a rich user experience that's been offered

[00:07:10] that was never possible before.

[00:07:12] Every time the experience becomes richer or something as disruptive

[00:07:17] as AI comes in and starts to generate 10 times,

[00:07:20] 100 times as much access to these services that are sort of combined

[00:07:27] in order to give one user experience.

[00:07:29] That actually increases the load on the backend database

[00:07:33] and that's where Cobroach sits.

[00:07:34] So think Cobroach is a disruptive player

[00:07:38] that can scale because of its truly distributed nature

[00:07:41] as large as it needs to be to satisfy these increasingly complex

[00:07:47] and rich sort of user experiences for all of the different platforms

[00:07:52] and services that you use in the course of a day

[00:07:55] which these days is I think probably difficult to catalog.

[00:07:59] And then the final one, and this is I think the most interesting

[00:08:02] and differentiated is just the reality of a modern business

[00:08:08] which has usually little difficulty appealing to people in Europe,

[00:08:14] people in the United States, people in Asia.

[00:08:16] So how do you juggle all of the requirements of such a business?

[00:08:21] You need to have a low latency interaction

[00:08:24] so that kind of argues you need to keep an Asian user's data

[00:08:27] somewhere nearby.

[00:08:29] You don't want to jump halfway across the world

[00:08:31] because those latencies can actually become quite noticeable

[00:08:34] to the end user.

[00:08:36] You also want to work around the increasingly complex requirements

[00:08:42] around data sovereignty.

[00:08:44] Where is the user's data domiciled?

[00:08:46] In other words, where is it located?

[00:08:47] Is it within their legal jurisdiction?

[00:08:49] And there's a number of different industries

[00:08:52] where that actually has become regulated

[00:08:54] and there is laws from the countries

[00:08:57] and various mandates and recommendations.

[00:09:01] So there's all different levels of them.

[00:09:02] But what Cobroach does with this distributed architecture

[00:09:05] is it's actually able to recognize that the location of data,

[00:09:10] the locality of it, is a very important dimension

[00:09:13] that needs to be used to restrict access in some cases,

[00:09:17] to guarantee that the data is never leaving

[00:09:20] a particular legal jurisdiction

[00:09:22] and to provide sort of underlying capability in the database

[00:09:26] to make sure that the information that you need

[00:09:29] is close to the user that needs it

[00:09:31] so that you can have really great user experiences

[00:09:34] with low latency.

[00:09:35] So those are just the kind of headliners,

[00:09:38] but there's a number of sort of supporting things

[00:09:41] that our customers are looking for.

[00:09:43] And I think chief among them is sort of freedom.

[00:09:46] A lot of the old school technologies

[00:09:49] really limit you geographically

[00:09:51] and to a single facility, maybe it was a backup somewhere.

[00:09:54] So Cobroach kind of busts through that.

[00:09:56] But increasingly, what companies

[00:09:59] at sort of the high end of the market,

[00:10:00] so think about big enterprises,

[00:10:02] are really concerned with is,

[00:10:04] what's their vendor concentration risk?

[00:10:06] In other words,

[00:10:08] did they wanna get fully in bed with one cloud provider

[00:10:10] because there's never been more

[00:10:14] that you're relying on one of these providers

[00:10:17] to offer or to support.

[00:10:20] And so that can create a really significant dependence.

[00:10:25] And Cobroach actually allows our customers

[00:10:28] to build their businesses kind of adgnostically

[00:10:31] above the public cloud.

[00:10:32] So that you're really using the best

[00:10:34] of what the multiple cloud providers out there are offering.

[00:10:37] You're sort of pulling them together

[00:10:40] to build a stronger business

[00:10:41] than anyone on their own could provide for you.

[00:10:44] And in fact, that also allows you to grapple

[00:10:48] with the reality of, okay, many of these big customers,

[00:10:50] they have their own data centers still.

[00:10:52] They want to use the public cloud,

[00:10:53] but they wanna have the option to get off of public cloud.

[00:10:56] And if they're using one particular

[00:10:57] public cloud provider,

[00:10:58] they want the option to have

[00:10:59] very seamless portability to others.

[00:11:02] So all of that is really where Cobroach has

[00:11:05] a significant strategic differentiation

[00:11:09] from our very real competitors

[00:11:12] in the sort of legacy space like Oracle

[00:11:16] and in the future state, current and future state,

[00:11:20] which is the big cloud providers like GCP, AWS and Azure.

[00:11:26] And I also say over the last six years,

[00:11:28] since we spoke,

[00:11:29] we've seen so many big changes in the world.

[00:11:31] It was a very different place back in 2018.

[00:11:34] And as that moved towards multi-cloud

[00:11:36] and hybrid cloud environments

[00:11:38] has increasingly become the norm,

[00:11:41] how do you at CobroachDB facilitate

[00:11:43] that transition for businesses?

[00:11:45] And what would you say makes it uniquely suited

[00:11:48] to thrive in such setups?

[00:11:50] Well, when we did this really interesting piece

[00:11:53] of research, which you can find by searching for it

[00:11:56] on Google, let's say, about it's

[00:11:58] the Cobroach's multi-cloud report.

[00:12:01] And we really went to a laundry list

[00:12:05] of the world's biggest companies

[00:12:06] and quiz them about sort of where they are

[00:12:09] in terms of their usage of public cloud and hybrid.

[00:12:13] What we found was that virtually

[00:12:15] all of these big companies are already multi-cloud.

[00:12:18] That was an interesting finding

[00:12:19] because it maybe was a little surprising.

[00:12:22] We didn't think quite that percentage would be.

[00:12:25] They're not all actively running their use cases

[00:12:28] across multiple clouds at once.

[00:12:30] That's sort of future state where people want to move.

[00:12:33] The reality though for these businesses is that

[00:12:35] through M&A or through just sort of organic internal interest

[00:12:42] they've actually already embraced multiple cloud vendors

[00:12:45] and many of them, as I mentioned in my previous comments

[00:12:48] are already have long been using

[00:12:51] their own private data centers.

[00:12:52] So when they think about like just the nature

[00:12:55] of their business in 2024,

[00:12:57] most of these big companies

[00:12:59] they're spread all over the place.

[00:13:01] So what they see top-roach DB is really providing

[00:13:04] and this is an essential differentiator

[00:13:10] that we bring to market is this consolidation point

[00:13:13] across all of those different environments

[00:13:15] where they're necessarily,

[00:13:21] I guess running their applications and services

[00:13:25] and data storage, it's called like databases.

[00:13:32] Sometimes for legacy information,

[00:13:33] sometimes for regulatory purposes

[00:13:34] they have to stay where they are on a mainframe.

[00:13:37] You want them to have a point that brings all this together

[00:13:39] at a database that's able to run in the private data centers

[00:13:42] that's able to actually straddle private and public.

[00:13:46] Maybe that's temporary as you're phasing things out

[00:13:50] in the private data centers

[00:13:52] and then you're gonna want to move everything

[00:13:54] into the public cloud.

[00:13:55] And the question is when you're making that movement

[00:13:57] of your applications and services

[00:13:59] and the really critical data sources,

[00:14:02] does the underlying database in this case, Cockroach DB

[00:14:06] make that seamless?

[00:14:07] Make it so that there's no downtime as you're moving.

[00:14:10] So there's no gravity that makes it so

[00:14:12] that it's impractical to move.

[00:14:14] And that's really what Cockroach DB is facilitating

[00:14:20] for these businesses.

[00:14:21] It's kind of like the central coordination

[00:14:26] and consolidation point where you can run flexibly

[00:14:30] in existing environments and all the new environments

[00:14:33] and you're not trapped in any particular decision.

[00:14:36] There's no sort of one-way doors.

[00:14:38] To use Cockroach in AWS as a starting point

[00:14:41] is not going to prevent you from moving to GCP

[00:14:45] or from moving to a situation

[00:14:47] where you're actively replicating across GCP, Azure and AWS.

[00:14:52] I was actually a really interesting place to be

[00:14:56] in terms of being able to survive a systemic cloud outage

[00:15:01] from one of those vendors.

[00:15:03] Well, you still got the other two.

[00:15:05] So you build a business that's as strong as any two

[00:15:07] out of all three of the major cloud players

[00:15:10] at least in the United States.

[00:15:11] Like that's a pretty fundamentally interesting place to be

[00:15:16] as one of the world's big businesses

[00:15:18] where you're not beholden to any one of these players

[00:15:21] but you're using all three to build something better

[00:15:23] than you could possibly build with any single one.

[00:15:28] And I think also with the growing concerns around things

[00:15:31] like cloud provider lock-in and the potential

[00:15:33] for excessive consolidation of power

[00:15:36] among these trillion dollar tech companies

[00:15:38] that we all know about,

[00:15:40] how do you at Cockroach DB empower enterprise

[00:15:43] to better maintain flexibility

[00:15:45] and avoid some of those pitfalls too?

[00:15:49] Oh, it's all about that portability promise.

[00:15:52] Yes.

[00:15:53] So Cockroach has grown up with the reality

[00:15:58] that the customers that select us

[00:16:00] are going to run us in many, many different environments.

[00:16:03] So that's a steeper learning curve for us

[00:16:06] and for the evolution of the software, right?

[00:16:08] Because it's much easier if you say,

[00:16:10] hey, we're just building this to run in AWS.

[00:16:12] And that's how AWS builds its database products.

[00:16:15] And that's a much simpler territory in which to build

[00:16:21] because there's fewer potential dimensions of variability.

[00:16:27] But once you say we're gonna run

[00:16:28] in all three different clouds,

[00:16:29] you start to realize very quickly

[00:16:30] that the three different clouds

[00:16:31] are pretty different in some fundamental way.

[00:16:34] They have different ecosystems of services.

[00:16:36] They have different idiosyncrasies.

[00:16:38] They have different kinds of reporting tools,

[00:16:40] different sorts of billing systems.

[00:16:41] And so you have to sort of generalize

[00:16:43] and build on top of this.

[00:16:45] Then when you add in the very necessary,

[00:16:48] but sort of infinitely complex reality

[00:16:50] of the private data centers,

[00:16:52] all the different things that have evolved

[00:16:54] over many decades across many different companies

[00:16:57] and industries that also would like to,

[00:17:01] you know, to this point about consolidation,

[00:17:03] Cockroach needs to run everywhere.

[00:17:05] But it may not stay running everywhere

[00:17:07] because there may be some transition points

[00:17:09] like an intermediate stages in a modernization story.

[00:17:12] But Cockroach can be there from the state,

[00:17:14] you know, state A through B through C through D

[00:17:19] in order to give you the state E, right?

[00:17:21] And that's what these businesses are looking for.

[00:17:23] That's what they fundamentally need.

[00:17:25] It's like, it's not like we can just flip a switch

[00:17:29] and have a, you know, the crown jewels of the business,

[00:17:32] right, where all of our most important customers,

[00:17:38] you know, maybe even, you know,

[00:17:40] we have customers that are, you know,

[00:17:42] holding the deposits for a significant amount of,

[00:17:46] you know, banking clients across the world.

[00:17:50] That's not something that you can lift and shift

[00:17:53] into a totally new environment.

[00:17:55] There's many intermediate steps.

[00:17:57] And so that's what our customers rely on us to provide.

[00:18:00] Like a system that works competently everywhere.

[00:18:03] And the beauty of that approach

[00:18:06] is that that gives you a,

[00:18:11] like a foundation upon which you have cloud portability

[00:18:15] inherent in your new architecture.

[00:18:17] You never, you weren't relying on one vendor

[00:18:20] and their, you know, very prescriptive

[00:18:22] and sort of idiosyncratic environment

[00:18:26] as the sort of starting point.

[00:18:29] Instead, you actually just because you had to move

[00:18:32] from hybrid to the public cloud,

[00:18:35] you've actually established yourself as,

[00:18:37] I like the way I like to say it is you've built your,

[00:18:40] your modernized business above the clouds, right?

[00:18:44] You're not in the stormy clouds.

[00:18:46] You're kind of above them.

[00:18:47] And so where one cloud might be stormy and having problems,

[00:18:50] the other clouds are clear and, you know,

[00:18:54] it's kind of a tortured metaphor.

[00:18:56] But like the goal here is really

[00:18:57] you're poking your head up above the clouds

[00:19:00] and you're sort of operating from a position

[00:19:03] of considerable strength.

[00:19:04] And I think, you know, to that idea

[00:19:07] of vendor concentration risk,

[00:19:09] this is very pertinent to the world of databases

[00:19:12] because most of these big businesses out there

[00:19:16] are of course, you know, decades long customers

[00:19:20] of the legacy database technology.

[00:19:22] And those vendors have, you know,

[00:19:25] in some cases been somewhat monopolistic

[00:19:28] and in their respective spheres in the past

[00:19:30] and haven't always had good customer relationships.

[00:19:33] And that actually really exacerbated

[00:19:36] the sort of institutional concern over how much,

[00:19:41] how many eggs are we putting into one vendor basket?

[00:19:44] What's interesting about the world we are in now

[00:19:46] is that that concentration has increased dramatically

[00:19:49] from many companies.

[00:19:50] And it kind of gone all in on one cloud provider

[00:19:53] because now this cloud provider

[00:19:53] is not just providing a database or databases.

[00:19:57] They're providing a, you know,

[00:20:00] of the lion's share of a corporate IT budget.

[00:20:04] I hundreds of millions of dollars in some cases a year.

[00:20:07] So this is considerably larger

[00:20:10] than the vendor concentration risk

[00:20:11] people used to worry about

[00:20:12] with a single say database provider.

[00:20:15] So that, you know, is definitely something

[00:20:18] that's hung these big companies radars to avoid

[00:20:22] and these, I think trillion dollar companies

[00:20:28] that look like they're gonna become

[00:20:29] 10 trillion dollar companies are,

[00:20:31] they offer a lot of advantages.

[00:20:33] But I think the businesses that kind of surfed

[00:20:35] this tsunami that's coming our way,

[00:20:37] especially with AI being added into the mix

[00:20:40] are the ones that are going to use the best aspects

[00:20:43] of all of these cloud vendors, you know, in aggregate

[00:20:47] to build something stronger

[00:20:48] than you can get with any single one.

[00:20:50] And of course at the moment there's so much

[00:20:52] economic uncertainty, global conflict,

[00:20:55] businesses are challenged with doing more with less

[00:20:58] and is a low scrutiny under every tech project

[00:21:01] for how much business value it delivers

[00:21:03] and ROI it delivers.

[00:21:05] And the database management market is also undergoing

[00:21:08] significant changes as a result

[00:21:10] with CIOs actively seeking ways now to cut costs

[00:21:13] and enhance efficiency and all that cool IT stuff.

[00:21:17] So how are you at Cockroach Labs

[00:21:19] navigating that shifting landscape?

[00:21:21] And are there any advantages

[00:21:23] you see offering businesses looking to adapt

[00:21:25] to this uncertain macro environment that we find ourselves in?

[00:21:30] That's, you know, it's a great question

[00:21:31] and there's sort of two ways to approach it.

[00:21:35] The kind of most obvious way is that

[00:21:38] we are constantly innovating to make the database

[00:21:42] less expensive or sort of a unit of work.

[00:21:45] Right.

[00:21:45] In other words, in a year from now

[00:21:49] the number of VCPUs that are running Cockroach

[00:21:53] in order to satisfy, let's call it like a million users

[00:21:58] on a particular service,

[00:22:00] we want to make that less expensive a year from now.

[00:22:04] Just by innovation, cutting costs,

[00:22:07] making the code more efficient and faster,

[00:22:10] that sort of thing.

[00:22:11] So we're always working on that.

[00:22:12] And one of the key pieces of technology we created

[00:22:15] recently is something called multi-region

[00:22:17] or multi-tenancy.

[00:22:19] So this is kind of like virtualization of databases

[00:22:22] because it turns out that

[00:22:24] because you never want your database to become

[00:22:27] so overburdened that it starts to be unable

[00:22:31] to keep up with a workload,

[00:22:33] people will typically over,

[00:22:37] they create excess capacity in the database system

[00:22:40] so they have a huge buffer.

[00:22:41] And it turns out that buffer is further exacerbated

[00:22:45] in terms of sort of what looks like a lot of overcapacity

[00:22:50] because businesses have natural peaks and troughs

[00:22:54] in terms of their demand.

[00:22:57] Sometimes that it can be really severe.

[00:22:59] Like your peak can be 10 times larger

[00:23:03] like a Black Friday type situation

[00:23:05] if you're retail as one example.

[00:23:08] Then it would normally be even at the high point

[00:23:11] of say a normal week.

[00:23:13] So you can have those sort of periodic events,

[00:23:15] other times it's just a naturally sort of bursty workload

[00:23:18] in terms of how things come in.

[00:23:20] Something like virtualization in the database,

[00:23:23] it recognizes that the average production database cluster

[00:23:27] is significantly underutilized.

[00:23:29] Sometimes it's so important that yeah,

[00:23:32] it's such a mission critical use case

[00:23:34] with so much value behind it

[00:23:36] that that's just the way you wanna run your database

[00:23:38] very safely, huge margins of safety.

[00:23:41] Other times you wanna say,

[00:23:43] hey we've got a whole bunch of use cases.

[00:23:45] If we can virtualize the database

[00:23:47] and run those with one sort of pool of physical resources

[00:23:51] that we can realize considerably better efficiency.

[00:23:55] Like we can drastically reduce the cost of databases.

[00:23:59] So that's a feature that we brought to market

[00:24:01] in our cloud platform.

[00:24:03] And that's actually already allowing customers

[00:24:05] to realize radically better total cost of ownership.

[00:24:09] That kind of brings me to the second point.

[00:24:10] I think this is the bigger one

[00:24:12] and that sort of take a step back and say,

[00:24:15] what's actually required to run these use cases?

[00:24:18] There's all these different costs, right?

[00:24:19] Cost of the cloud computing,

[00:24:20] the cost of the database, the software,

[00:24:23] the human labor that's required for example,

[00:24:25] to build the use case and to maintain it

[00:24:27] and to iterate the features.

[00:24:29] So these are with the software engineers

[00:24:31] that are doing your application use case.

[00:24:33] But there's also human labor

[00:24:35] that's required to run the database,

[00:24:36] the DBAs, the consultants that do performance optimization.

[00:24:41] And what is true and part of the reason

[00:24:45] that we built Cockroach is that those costs,

[00:24:48] especially the human costs of trying to run a database

[00:24:52] that is struggling with scale for example,

[00:24:55] or struggling because there's too many outages,

[00:24:59] those human costs can become prohibitively expensive.

[00:25:02] Like in the early odds where many internet companies

[00:25:07] were getting their start, think Airbnb

[00:25:10] and Facebook and things like that.

[00:25:14] These companies ended up having to build their own databases

[00:25:16] just because databases cannot scale.

[00:25:18] And let me tell you,

[00:25:19] it is very, very expensive to build your own database.

[00:25:21] It's a many year project,

[00:25:22] many centuries of engineering labor,

[00:25:27] like in aggregate,

[00:25:28] go into these buildings something like this.

[00:25:30] So the costs of building your own database are staggering.

[00:25:34] And in many companies have been forced

[00:25:35] to do that because of scale.

[00:25:36] So when you take a step back

[00:25:38] and you see what Cockroach is providing,

[00:25:40] it's a solution to these things

[00:25:41] where you don't have to either constantly be struggling

[00:25:45] with all of the displaced productivity

[00:25:47] and human costs there.

[00:25:49] But you actually are able to solve problems

[00:25:54] that have forced companies in the past

[00:25:56] to actually become specialists in building databases,

[00:26:00] which is a, you know,

[00:26:01] that often comes at a significant cost.

[00:26:04] I mean we're talking about

[00:26:05] tens or hundreds of millions of dollars.

[00:26:08] And so what's the cost of Cockroach compared to that?

[00:26:10] It's actually minuscule

[00:26:11] because we're bringing that to many companies at once.

[00:26:14] All right, so the secret to really cutting costs going forward

[00:26:18] is to use better infrastructure.

[00:26:21] All right, the costs for example

[00:26:22] of having a cloud regional outage

[00:26:25] can be quite high to, you know,

[00:26:27] your business depending on what it is.

[00:26:29] If that's not, if that's a risk

[00:26:32] that is significantly attenuated

[00:26:34] by Cockroach's resilience capabilities,

[00:26:36] then you've just, you know,

[00:26:38] at least probabilistically saved a huge amount of cost,

[00:26:42] you know, that could have come to your brand

[00:26:43] or it could have just been dollars and cents

[00:26:45] in terms of lost business due to an outage

[00:26:47] and certainly tons of engineering productivity

[00:26:50] and frustration.

[00:26:52] So those are all, I think those are the real efficiencies

[00:26:56] that can be gained by using better technology

[00:26:58] and that's really what we've brought to market.

[00:27:01] So as we all observe this possible decline

[00:27:04] in mainframe use due to talent shortages

[00:27:06] and advancing competing technologies,

[00:27:09] what role do you see cloud-maker solutions

[00:27:11] like Cockroach DB playing

[00:27:13] in helping organizations transition smoothly

[00:27:16] to that tech utopia, which is of course

[00:27:19] modern and scalable architectures

[00:27:21] because it doesn't have to be just a pipe dream,

[00:27:24] is it? You do this all the time

[00:27:26] and when you're helping people.

[00:27:28] Well, you know, I'd say that it was a pipe dream

[00:27:30] until fairly recently.

[00:27:32] I think we've been a key player

[00:27:34] and actually moving some pretty consequential use cases

[00:27:36] off some of these aging legacy technologies,

[00:27:39] which by the way are very good,

[00:27:40] which is why it's been such a difficult road

[00:27:43] to a credible successor technology.

[00:27:46] And that's really what we are.

[00:27:47] You know, what makes that possible?

[00:27:49] The reality is that there's almost nothing

[00:27:51] that's I think ever been created to my knowledge

[00:27:54] that is more resilient than a mainframe

[00:27:58] sitting in a single facility.

[00:28:00] So those provide seven nines of availability

[00:28:04] within a particular facility.

[00:28:07] The problem is that the facility itself

[00:28:09] can lose connectivity or power.

[00:28:13] And that's where Cockroach is bringing something

[00:28:15] that's a stronger guarantee for certain kinds of issues

[00:28:18] than what the mainframe can provide.

[00:28:21] It's kind of like you get something and you lose something.

[00:28:24] Mainframes are very great pieces of technology,

[00:28:27] but they're fundamentally monolithic

[00:28:30] and they were designed really with the expectation

[00:28:33] that the facility that it's running in

[00:28:37] would remain connected.

[00:28:38] And if it didn't,

[00:28:39] that was just something that you dealt with.

[00:28:41] Now people don't want to deal with that

[00:28:43] and there's technology like CockroachDB

[00:28:45] that provides a solution there.

[00:28:47] So that's a big part of it.

[00:28:49] But there's this tailwind now,

[00:28:50] which is first the interest in running

[00:28:55] in the public cloud and in modern ecosystems.

[00:28:57] So you want to get out of that private data center

[00:29:00] that's running the mainframe.

[00:29:02] You want to run your applications

[00:29:03] and richer ecosystems where you can use these LLMs

[00:29:07] and things like that.

[00:29:08] So that's just one example that's come up more recently,

[00:29:10] but also all the different services in the public cloud,

[00:29:13] you can build better things faster

[00:29:15] if you're able to use these rich ecosystems.

[00:29:17] So that's providing a tailwind,

[00:29:18] but also as you pointed out,

[00:29:20] the people that know how to program

[00:29:22] and maintain and run these mainframes

[00:29:24] are aging out of the workforce.

[00:29:25] This is very old technology

[00:29:26] and it's not considered the future by many.

[00:29:32] So that's actually also providing quite a bit of impetus

[00:29:35] around our kinds of solutions.

[00:29:37] And so Cockroach is providing

[00:29:39] a slightly different flavor of resilience

[00:29:43] that is actually very important in the world of the cloud,

[00:29:46] I think the world that we've all entered.

[00:29:48] And so I think it's a credible successor technology

[00:29:53] for databases beyond what mainframes can provide

[00:29:57] in the current moment.

[00:29:58] And so that's really helping these businesses say,

[00:30:02] okay, what are we gonna bet on for the next 10 years?

[00:30:05] Like how are we gonna build our systems?

[00:30:07] We wanna find something that is,

[00:30:09] that's going to last at least that long

[00:30:11] because it's a very expensive transition.

[00:30:13] And also I wanted to highlight today

[00:30:15] is the concept of success disasters,

[00:30:18] which highlights the challenges businesses face

[00:30:21] when sudden spikes in online demand

[00:30:23] exited their data management capabilities

[00:30:25] and we've all seen what happens there

[00:30:27] and how it can take an entire store or website

[00:30:30] or business completely offline,

[00:30:31] bring down the shutters to their business.

[00:30:33] So can you share any examples of how Cockroach DB

[00:30:36] has maybe helped companies overcome challenges like that

[00:30:39] and any lessons that can be learned from these experiences

[00:30:43] because when things like this happened on,

[00:30:45] I don't know, like a big event like Black Friday,

[00:30:47] it's game over for organizations, isn't it?

[00:30:50] Yeah, I mean, losing out on that one day

[00:30:52] because your system crashes is very painful.

[00:30:56] And I should think that risk of success disasters

[00:30:59] is increasing as the sort of qualitative experiences

[00:31:04] can become far better overnight with things like AI

[00:31:08] being blended into a particular product experience.

[00:31:11] And so you can end up with many, many people

[00:31:12] trying it out like viral word of mouth.

[00:31:14] And so you can have, I think, unsurprising amounts

[00:31:18] of traffic that show up just because

[00:31:20] things are changing very rapidly, right?

[00:31:23] Certainly there's such a profusion of these things

[00:31:26] and whatever really hits home,

[00:31:30] I think that can create a more cute success disaster

[00:31:33] than was really ever possible before.

[00:31:35] There's just slower word of mouth, now that's fast,

[00:31:37] but I think you can create a game changer demo

[00:31:42] in probably almost any kind of use case

[00:31:44] that you can think of.

[00:31:46] We're gonna see a lot of those over the coming years

[00:31:48] and those can all lead to big success disasters.

[00:31:51] A scalable architecture like Cockroach

[00:31:53] makes those much easier to survive.

[00:31:55] You can add those nodes in real time

[00:31:59] in our cloud platform, if you're using

[00:32:04] what we call serverless capability there,

[00:32:06] you can scale up, you don't even have to plan for that.

[00:32:09] It's sort of auto-scales in real time.

[00:32:11] So that's a really great way

[00:32:15] to avoid a success disaster.

[00:32:17] You just kind of, you get that for free almost.

[00:32:20] The biggest kind of lesson learned here is just

[00:32:22] you've got to be testing for these things.

[00:32:25] You've got to identify what that max capacity

[00:32:30] that you have a reasonable expectation for,

[00:32:33] like what are your ambitions?

[00:32:34] Like what is your hope

[00:32:35] for what could happen on Black Friday,

[00:32:37] that sort of thing?

[00:32:37] You definitely need to test and plan for those things

[00:32:40] and then you also need to architect your system

[00:32:42] so that it fails fairly gracefully

[00:32:44] when you exceed what that maximum was

[00:32:46] because that's always a possibility.

[00:32:48] And so what does that mean?

[00:32:49] It means that your whole system doesn't go down.

[00:32:51] You can still process your max number of requests

[00:32:54] but everyone else gets put into a queue.

[00:32:56] That's not a crazy concept

[00:32:57] but most people don't test at that point

[00:32:59] and so their whole system can crash

[00:33:00] if they go beyond a certain level of scale.

[00:33:03] So part of it is the database

[00:33:05] and we offer those,

[00:33:07] that sort of foundational scaling capability.

[00:33:09] And part of it is, okay,

[00:33:11] what's the application that's using the database

[00:33:13] and how we tested it at those extreme scales?

[00:33:17] And if we were to look ahead a little further

[00:33:19] into the future,

[00:33:21] how do you envision the role of distributed SQL databases

[00:33:24] evolving, especially with the increasing integration

[00:33:27] of January to May,

[00:33:28] which is all everyone's talking about

[00:33:30] and that gets integrated into business operations.

[00:33:32] And with all these changes happening

[00:33:34] in what steps is Cochrose Labs taking

[00:33:37] to remain at the forefront of this evolution

[00:33:39] because it just feels like they're moving

[00:33:41] at such breathtaking speed, right?

[00:33:43] I think a big part of this

[00:33:45] is that data intensity angle.

[00:33:49] When you think about what backend databases

[00:33:55] have to contend with,

[00:33:57] it all comes down to how many agents

[00:34:00] are kind of knocking on the door.

[00:34:02] It used to be how many people had desktop computers.

[00:34:05] Then mobile came along

[00:34:07] and now you went from like 300 million desktop computers

[00:34:09] to 10 billion mobile devices,

[00:34:13] maybe quite a bit more.

[00:34:14] And then you had,

[00:34:17] now I think what we're gonna see

[00:34:18] is sort of virtual agent proliferation.

[00:34:21] So you go to 100 billion into a trillion, right?

[00:34:25] Things that are out there hitting APIs

[00:34:27] will be more APIs and so forth,

[00:34:29] but the popular ones are gonna massively,

[00:34:33] they're gonna see massive ways of consumption.

[00:34:36] And so that's going to make

[00:34:41] operational data, big data.

[00:34:44] In the past when you heard the term big data,

[00:34:46] it usually meant like everything you've ever collected

[00:34:48] on all your customers are putting into a data warehouse

[00:34:50] and trying to glean business intelligence from it.

[00:34:52] But we're now entering a world

[00:34:54] where the operational database, the systems of record,

[00:34:57] those are big data.

[00:34:58] And those need to get scaled

[00:35:00] to petabyte scales and so forth.

[00:35:02] So I just see that AI has the potential

[00:35:05] to make that the reality across every industry.

[00:35:09] Just because it leads into this idea

[00:35:12] of a richer user experience,

[00:35:13] that we have one customer that fulfills orders.

[00:35:18] It's a sort of a logistics company.

[00:35:20] And they have estimated that the average number

[00:35:25] of interactions with the database

[00:35:28] to satisfy one customer order can be as many as 5,000.

[00:35:33] I was gonna tell you in like the recent past,

[00:35:36] that would be an impossibility.

[00:35:40] Like that'd be inconceivable, 5,000 different access

[00:35:44] to database, you try to make it one access to the database,

[00:35:46] like one transaction or maybe several.

[00:35:51] That would be sort of the max,

[00:35:52] otherwise you'd overload these systems.

[00:35:54] So like that is only going to be a trend that continues.

[00:35:58] So we actually see that distributed SQL is,

[00:36:01] this is what we've bet our company on even 10 years ago,

[00:36:05] that the trend here in data intensivity

[00:36:07] is gonna make every operational use case

[00:36:10] potentially need what people used to call big data.

[00:36:13] But now it's kind of like the goal posts are moving.

[00:36:16] So distributed SQL, that concept of a database

[00:36:21] no longer being monolithic,

[00:36:22] but really just expanding according to the needs

[00:36:25] of the underlying use case.

[00:36:26] That of course is the right way to build.

[00:36:29] Not just in the database,

[00:36:30] the whole technology stack has to be built that way.

[00:36:32] But the database can be often the hardest,

[00:36:35] most sort of foundational element

[00:36:36] that everything else is sort of relying on

[00:36:39] so that they also get that scalable capacity.

[00:36:42] A question I've got to ask before I let you go today

[00:36:44] is the database management market

[00:36:46] is probably doing something like $80 billion and growing.

[00:36:50] So any future developments

[00:36:52] that you might anticipate in the industry

[00:36:54] or things that excite you, anything you're seeing

[00:36:56] and also how again,

[00:36:58] how your cockroach labs are maybe prepared

[00:37:00] to meet some of those emerging needs and opportunities.

[00:37:03] You know, this is a fast-growing market.

[00:37:05] The expectation is that in 10 years,

[00:37:08] it'll be a quarter of a trillion dollars.

[00:37:10] Market.

[00:37:11] And again, you're right in the previous question

[00:37:16] to really point out generative AI

[00:37:18] and that's just maybe the tip of the iceberg.

[00:37:22] But this is definitely the apocleship

[00:37:26] that will launch a million new use cases

[00:37:30] and certainly the refactoring of existing ones.

[00:37:33] So we actually see that the database market grows

[00:37:37] in proportion to those new starts, right?

[00:37:39] And to those refactoring.

[00:37:41] And so the estimates,

[00:37:45] the extrapolation of like the compound annual growth rate

[00:37:48] and database applications and so forth

[00:37:50] might be a severe underroll statement.

[00:37:53] And so what do we fundamentally have to do?

[00:37:55] Actually, I believe that the world ahead,

[00:37:58] AI is making the barriers to entry extremely inexpensive.

[00:38:03] Right?

[00:38:04] To build a service that looks like another service

[00:38:06] is just going to get easier and faster

[00:38:08] and require less resources.

[00:38:11] And that's already been happening

[00:38:12] because you don't have to run your own data centers anymore.

[00:38:14] You don't have to provision in your own machine

[00:38:17] and build them and put them

[00:38:18] into some co-location facility.

[00:38:20] You don't have to run the database anymore

[00:38:23] or the message passing

[00:38:26] or the provisioning of servers and things.

[00:38:31] You can use serverless compute technologies

[00:38:35] like Lambda and so forth.

[00:38:37] So a small team can build what used to take

[00:38:40] hundreds of people to competently manage.

[00:38:44] That's only going to get faster and faster.

[00:38:46] It's already happening in real time.

[00:38:49] I think that we have like an opportunity

[00:38:52] based on where we play in the market

[00:38:54] to truly be serving the businesses

[00:38:57] that are going to win this environment.

[00:39:00] And it turns out, I mean, this is just my opinion here,

[00:39:03] but the brand that a business has

[00:39:09] and the data they have around their business,

[00:39:13] maybe operating for decades or more.

[00:39:17] And like when things become really inexpensive

[00:39:21] and there's this sort of this massive acceleration

[00:39:24] of evolution in use cases

[00:39:26] and it's just ever cheaper to build what already exists,

[00:39:30] that's where brands are going to be

[00:39:31] the deciding factor in success.

[00:39:34] What's that brand loyalty?

[00:39:35] What do people look to?

[00:39:36] What's the muscle memory?

[00:39:39] If a challenger comes along

[00:39:41] and they have a way better AI powered use case,

[00:39:43] those shave off from customers,

[00:39:45] but how long does it take

[00:39:46] the true kind of incumbent

[00:39:51] in that particular product category

[00:39:52] to duplicate those capabilities?

[00:39:54] Well, no longer than it took that startup to build them.

[00:39:57] And so like what cockroach plays in this market

[00:40:01] is we serve the world the biggest businesses

[00:40:03] that sort of hands down.

[00:40:05] Those are the ones that need our scale.

[00:40:07] Those are the ones that insist on our reliability,

[00:40:09] our resilience,

[00:40:10] because those are the brands that can't go down.

[00:40:13] They're that valuable.

[00:40:15] That many people rely on them.

[00:40:16] They're front page of the Wall Street Journal,

[00:40:17] something happens.

[00:40:19] So like those are already our customers.

[00:40:21] Those are gonna be the biggest beneficiaries

[00:40:23] of this AI boom

[00:40:24] because they can build whatever anyone else can build

[00:40:27] just as quickly.

[00:40:28] They might go a little more slowly.

[00:40:30] They'll have their innovators dilemmas,

[00:40:31] but it's becoming sort of,

[00:40:35] there's sort of a race to the bottom

[00:40:36] in terms of the cost to produce.

[00:40:38] So the brands are gonna be the telling factor here.

[00:40:41] It's just something that,

[00:40:44] we're paying a lot of attention to.

[00:40:47] And it would be something I would look to,

[00:40:52] pay a lot of attention to in the couple of years ahead

[00:40:55] to see how that dynamic plays out.

[00:40:57] Then six years ago,

[00:40:58] you shared that tech startup story

[00:41:00] of 3DX Googlers and cockroach labs.

[00:41:03] And here in 2020,

[00:41:04] all cockroach DB is a preferred solution

[00:41:06] for companies like Netflix,

[00:41:08] Bob's, Comcast, and as you said earlier,

[00:41:10] they're just a few of the ones that we're able to name.

[00:41:13] But as we come full circle,

[00:41:14] I'm gonna ask you to look back

[00:41:16] and reflect on that tech startup journey

[00:41:18] because I think none of us are able to achieve

[00:41:21] any degree of success without a little help along the way.

[00:41:23] So I'm curious, is there a particular person

[00:41:26] or someone you're grateful towards?

[00:41:27] Maybe they saw something in you,

[00:41:30] anything at all that maybe just helped you get you

[00:41:32] where you are today,

[00:41:33] that we could give that person a shout out.

[00:41:34] But who would it be?

[00:41:36] As you point out,

[00:41:37] success has many inputs to it.

[00:41:43] You don't get as far as I've gotten in my career

[00:41:48] without huge amounts of assistance.

[00:41:53] So there's everyone that's helped to build cockroach,

[00:41:56] but I just call out in particular,

[00:41:57] we've gotten extremely lucky in the advisors

[00:42:00] we've had along the way.

[00:42:02] And starting with one of our investors,

[00:42:05] his name is Peter Fenton,

[00:42:07] and almost call out Mike Volpe.

[00:42:08] They let our, respectively,

[00:42:10] our first and our second rounds in the business

[00:42:12] and they've both been on our board.

[00:42:14] So it's been nine and eight years, respectively.

[00:42:17] And they have kind of,

[00:42:20] they're sort of perched at the apex of the ecosystem,

[00:42:23] watching all of these different companies.

[00:42:26] One is from Benchmark,

[00:42:27] one is from Index Ventures.

[00:42:29] And so that,

[00:42:31] they're also kind of,

[00:42:35] storied industry veterans

[00:42:37] have seen many of these things come and go.

[00:42:39] And so I would just say that they're consistent support

[00:42:43] and it's not,

[00:42:46] it's a very challenging kind of support.

[00:42:49] It's very constructive,

[00:42:50] a lot of iteration and feedback

[00:42:52] and constructive challenge just.

[00:42:58] And that has been instrumental,

[00:43:02] I think in helping us guide our business in the evolution.

[00:43:07] Absolutely, love that big shout out to both Peter and Mike.

[00:43:10] I think it's so important to recognize these individuals

[00:43:13] that maybe not aware of just how much you appreciate

[00:43:16] the help that you've been given along the way.

[00:43:18] And for anyone listening,

[00:43:19] just wants to find out more information

[00:43:21] about any of the topics we discussed today.

[00:43:23] We did cover a lot there,

[00:43:24] but for anyone listening wanting to dig a little bit deeper,

[00:43:26] connect with you, connect with your team,

[00:43:28] just find out more information.

[00:43:29] Where would you like to point everyone?

[00:43:31] Well, the right place to go would be our

[00:43:35] webpage, so that'd be cockroachlabs.com.

[00:43:37] And then all of the juicy information and details

[00:43:42] and an increasing amount of documentation

[00:43:46] and case studies and things would be where I'd go

[00:43:49] to learn everything about Cockroach DB

[00:43:52] and also about the company

[00:43:53] if anyone's interested in joining.

[00:43:55] Well, I'll add the links to that

[00:43:56] so people can find you in Iceland as well

[00:43:58] as the report that we referenced earlier

[00:44:01] in our conversation too.

[00:44:03] And it was just great catching up with you.

[00:44:05] I mean, I love how you're disrupting the database market,

[00:44:10] industry discussions today around the shift to the cloud

[00:44:14] and gen AI success disasters as a result of spikes

[00:44:18] in online customer demand and so much more.

[00:44:20] But more than anything,

[00:44:21] it was just great catching up with you

[00:44:22] after six long years where the world has changed so much.

[00:44:25] We can't leave it another six years.

[00:44:27] Who knows what's gonna happen in that time,

[00:44:29] but more than anything,

[00:44:30] just thank you for coming back on today, Spencer.

[00:44:32] Well, we won't be in a Terminator dystopia.

[00:44:34] So I hope we can connect in six years again.

[00:44:38] So as we wrap up today's conversation with Spencer,

[00:44:41] I think it's clear that the path to innovation

[00:44:43] in database technology

[00:44:45] is not just about handling data more efficiently,

[00:44:48] but doing so in a way that aligns

[00:44:50] with the future of the business in the cloud.

[00:44:53] And Cockroach Labs stands at the forefront

[00:44:55] of this transformation.

[00:44:56] They're challenging the status quo

[00:44:58] and offering a glimpse into a future

[00:45:00] where businesses can truly scale without bounds

[00:45:04] and navigate that regulatory landscape with ease.

[00:45:07] And ultimately avoid those pitfalls

[00:45:10] of things like vendor locking.

[00:45:12] Hopefully we can get rid of that once and for all.

[00:45:14] But I think also this conversation

[00:45:17] begs an even bigger question.

[00:45:19] As more companies move towards a multi-cloud strategy,

[00:45:23] how will the dynamics of data management

[00:45:25] and cloud portability,

[00:45:27] how will they evolve?

[00:45:28] And what role will technologies like Cockroach DB

[00:45:32] play in ensuring that businesses are not just surviving,

[00:45:35] but thriving in this new environment.

[00:45:37] But over to you.

[00:45:38] You've got more experience in this world than I do.

[00:45:40] So please share your thoughts.

[00:45:42] What are you doing about this?

[00:45:43] Please email me now techblogwriter.outlook.com,

[00:45:47] Twitter, LinkedIn, Instagram, just at Neil Seahues.

[00:45:50] And if you enjoyed today's episode,

[00:45:52] I'll be back again tomorrow.

[00:45:53] We'll explore a completely different topic

[00:45:55] about how technology is transforming our life,

[00:45:58] work and the world.

[00:46:00] But more than anything,

[00:46:01] just thank you for listening as always

[00:46:02] and until next time.

[00:46:05] Don't be a stranger.