Adam Jacob

And it all has to happen fast enough that it feels basically like real time. That was a very difficult scaling challenge, right? Because when you look at traditional database technologies, they're all fantastic on some different angle. but they're not quite the right fit.

And it all has to happen fast enough that it feels basically like real time. That was a very difficult scaling challenge, right? Because when you look at traditional database technologies, they're all fantastic on some different angle. but they're not quite the right fit.

And it all has to happen fast enough that it feels basically like real time. That was a very difficult scaling challenge, right? Because when you look at traditional database technologies, they're all fantastic on some different angle. but they're not quite the right fit.

And so to get that interactivity and to get the shape of the system correct, we had to do a lot of work basically designing our own in-memory database. And it's super cool. I'm very proud of it. And also, whoa. Those have been a lot of the big sort of scale challenges. You mentioned sort of building for scale.

And so to get that interactivity and to get the shape of the system correct, we had to do a lot of work basically designing our own in-memory database. And it's super cool. I'm very proud of it. And also, whoa. Those have been a lot of the big sort of scale challenges. You mentioned sort of building for scale.

And so to get that interactivity and to get the shape of the system correct, we had to do a lot of work basically designing our own in-memory database. And it's super cool. I'm very proud of it. And also, whoa. Those have been a lot of the big sort of scale challenges. You mentioned sort of building for scale.

And I think in general, the way you build for scale is to try to think about how those distributed systems will work under pressure, and more importantly, how they fail under pressure. If the system's going to break, does it break in predictable ways? And how do you know what they are?

And I think in general, the way you build for scale is to try to think about how those distributed systems will work under pressure, and more importantly, how they fail under pressure. If the system's going to break, does it break in predictable ways? And how do you know what they are?

And I think in general, the way you build for scale is to try to think about how those distributed systems will work under pressure, and more importantly, how they fail under pressure. If the system's going to break, does it break in predictable ways? And how do you know what they are?

If you can understand what those boundaries are between the different components of the system, then you can start to understand predictably, here's the couple of different ways that the system could possibly fail. And then you can look at any individual failure and start to assess what you're going to do about it. There's a great story. I don't know if it's apocryphal or not.

If you can understand what those boundaries are between the different components of the system, then you can start to understand predictably, here's the couple of different ways that the system could possibly fail. And then you can look at any individual failure and start to assess what you're going to do about it. There's a great story. I don't know if it's apocryphal or not.

If you can understand what those boundaries are between the different components of the system, then you can start to understand predictably, here's the couple of different ways that the system could possibly fail. And then you can look at any individual failure and start to assess what you're going to do about it. There's a great story. I don't know if it's apocryphal or not.

I think it was Brian Kernaghan and Ken Thompson were pair programming in the early days of Unix and they were trying to fix a bug. And I think Kernaghan was at the keyboard and he was like attaching, attacking this bug. Ken Thompson was standing behind him and he was getting all frustrated.

I think it was Brian Kernaghan and Ken Thompson were pair programming in the early days of Unix and they were trying to fix a bug. And I think Kernaghan was at the keyboard and he was like attaching, attacking this bug. Ken Thompson was standing behind him and he was getting all frustrated.

I think it was Brian Kernaghan and Ken Thompson were pair programming in the early days of Unix and they were trying to fix a bug. And I think Kernaghan was at the keyboard and he was like attaching, attacking this bug. Ken Thompson was standing behind him and he was getting all frustrated.

And then Thompson had his like eyes closed for a couple of minutes and then opened his eyes and was like, maybe open this file and go to this line number. And there was the bug.

And then Thompson had his like eyes closed for a couple of minutes and then opened his eyes and was like, maybe open this file and go to this line number. And there was the bug.

And then Thompson had his like eyes closed for a couple of minutes and then opened his eyes and was like, maybe open this file and go to this line number. And there was the bug.

As I've gotten older and my career's gotten longer, I really have learned to understand what I think he was doing in that moment, which is he was thinking through the boundaries of how the software worked, of the different ways that the different pieces of software interacted with each other, and how could those things possibly interact in order to create the outcome that you're seeing that's wrong.

As I've gotten older and my career's gotten longer, I really have learned to understand what I think he was doing in that moment, which is he was thinking through the boundaries of how the software worked, of the different ways that the different pieces of software interacted with each other, and how could those things possibly interact in order to create the outcome that you're seeing that's wrong.