text stringlengths 0 2.67k |
|---|
**Justin Garrison:** In a lot of ways, people are making the right decision to simplify their view of the system... But there's so many other ways that people are trying to view the system, that like "One table for me is good, because it's my table." But once someone else wants to access it, I'm like "Actually, now thi... |
**Pete Naylor:** Yeah. And it's like, there is never any free lunch in databases. So you look at it and you go "Well, you could stuff everything in one table, and maybe you think that's beneficial... But let me show you why that's costing you more." |
**Autumn Nash:** Yeah. The way that people think that one size fits all - databases are so complicated, because there's so many... It just goes back to the CAP theorem. You have to pick what's most important, and then not looking at the data and not having the context, assuming that you can give one magical concept to ... |
**Pete Naylor:** And I think that's the thing with relational databases, right? They're really, really cool, they're very powerful, very flexible... And I think it allows developers, people who maybe haven't really thought through exactly what it means to have those asset properties - not all of them really want to dig... |
\[57:54\] You could also take a relational database, and many of them now - there's a lot of different extensions for Postgres - and you could shard that out, and kind of build your own DynamoDB, in a way. Once again, there's no absolutes in any of this stuff... And for me, rather than getting wound around like new ter... |
**Justin Garrison:** SQLite. |
**Pete Naylor:** Just use SQL, right? Because, for one thing, it works from the whole transactional and all the way into analytics. Everybody learns it. Just - it can make things a lot easier. |
**Autumn Nash:** Do you think that you can scale a relational database and give it like the distributed -- like, same kind of architecture as like a relational, a non-relational database, and it can still be as performant? |
**Justin Garrison:** A not-Camry? Yeah... |
**Pete Naylor:** Yeah, so I think you can. You can take Postgres and if you just choose to make only those very simple queries, you can definitely do the same indexing, and you can split things out. What DynamoDB does that's still very, very special, even in the whole NoSQL space, where people took this sharded approac... |
**Autumn Nash:** Dynamo is a beast. |
**Pete Naylor:** Yeah. I mean, it's just like, here's this thing that if you have a certain simplified kind of a use case, you can just throw a load at it and every half hour it can double its ability to carry extra load, and double its ability to carry extra storage. And there's no limit. |
I remember going and working with some customers that were maybe using a MongoDB, or something... And MongoDB is interesting, because it started kind of in this NoSQL sharded approach, but then they really followed the developer side of things a lot, which - more power to them, right? They've had a lot of success with ... |
**Autumn Nash:** Well, I think it's also because they're one of the easiest databases to just throw some stuff in, and you don't have to have any background of like access modeling, or schemas, or anything. |
**Justin Garrison:** They should have called it YoloDB. |
**Autumn Nash:** It is! It is YoloDB! No, but honestly, it's one of those things where like sometimes I don't even think it's like picking the right tool. It's picking a tool that people that you employ can use, and MongoDB -- and they've also focused a lot on making like documentation and things friendly, for people t... |
**Pete Naylor:** It's a lesson, really, to me. The way that they documented things, the way they've taken the hearts of developers. Everybody should be paying attention to that. But I think over time -- |
**Autumn Nash:** That is one of their most underrated sales, is the way that they help people learn how to use their products. Everybody should steal that. Like, it is so good. |
**Pete Naylor:** But the flip side is it's kind of like "Well, we're going to give you these really nice features, that let you do these super-simple queries. And here you are, take this candy..."" And then if you actually start building around that and then you scale out, then it doesn't work quite as well with the wh... |
**Justin Garrison:** \[01:02:19.20\] One thing I find interesting in just infrastructure in general is... Like, that database and storage layer has typically stayed pretty - I don't wanna say closed source, but more proprietary than compute infrastructure. Getting a VM and running compute is pretty just generic, and it... |
**Autumn Nash:** Networking is really complicated. You're either good at that... |
**Justin Garrison:** Yeah, and there's no like MongoDB of networking. |
**Autumn Nash:** Because you can't like make that that simple. Like, one of the best parts of MongoDB is simplifying it so anybody can use it. And networking - I swear, if you're good at networking, you're good at it. And if you're everybody else, it's like watching Charlie Brown's teacher talk to you. |
**Justin Garrison:** Which is also funny, because think of think of every house that has a LAN. They don't have databases at their house, but they've all got LAN, and they've all got DNS, and they all somehow connected to this interconnect network of like wireless working place. Some people - they can learn that abilit... |
And then databases - there are the core open source, the Postgreses that are still like "Yeah, no, this is fundamental. You should be able to do this with all these extensions." But like a lot of it is still proprietary, and is still like "Nah, you're going to pay someone millions of dollars for that privilege." |
**Autumn Nash:** I think it's also because it's not sexy tech. |
**Justin Garrison:** It's so critical. |
**Pete Naylor:** \[unintelligible 01:03:59.02\] |
**Autumn Nash:** Well, I like databases, but how many people do you know wake up and they're like "I can't wait to become a database engineer"? And it's so abstracted half of the time... You know how we were talking about how like cloud has made infrastructure more abstracted, so people don't always have the context...... |
**Pete Naylor:** This kind of gets at something that's a little bit new for me. When I was on DynamoDB, there's no open source version of DynamoDB, right? There's no option there. |
**Justin Garrison:** Yeah. Same with Spanner. I mean, Spanner has Cockroach, or whatever... But there's a lot of -- like CosmosDB... Like, there's so many clouds that have this proprietary, big database that like you're just going to pay a bunch of money for. |
**Pete Naylor:** And I look to that as like "Man, there's no open source." And that's actually like -- it's good to give people that flexibility. There were reasons why people might have wanted to have like a DynamoDB-like thing on-prem. Didn't solve that for them. And I just think -- like, EnterpriseDB is really about... |
**Justin Garrison:** You get the experts. You are paid for the responsibility. |
**Pete Naylor:** You get the experts, and then they also do contribute back to the core project. It's like "What features? How do we do this?" and it's something that evolves over time. It's a really interesting balance to try and manage for... Because I can I can see it both ways, but you've really got to hit it just ... |
\[01:06:13.24\] So it's an interesting space... But databases are like -- the more I look at them, it comes back to storage. Recently I was looking at like - you make certain choices about replication within Postgres, for example... Or you could say "Well, I'm going to push this down and have this distributed storage, ... |
I had the opportunity recently to actually go in and see a data center, and actually touch servers, which I had not done in a really long time... |
**Autumn Nash:** Did you have flashbacks? |
**Pete Naylor:** Yes, I did. Many, many scary moments, like getting paged, going into a data center with a big rack full of drives, and like accidentally grabbing the wrong one... And it's like "Okay, this thing had a RAID volume that was okay if it lost one drive, and I needed to replace that one..." I pulled a differ... |
**Autumn Nash:** What did you go to the data center for? |
**Pete Naylor:** Oh, on that occasion? |
**Autumn Nash:** Yeah. |
**Pete Naylor:** I was just getting paged because one of the drives in this drive array had failed. |
**Autumn Nash:** No, this time that you went back. |
**Pete Naylor:** Oh, the new thing is we're actually kind of working on more of a holistic platform approach for Postgres. So kind of merging the transactional stuff, some of the analytics that's happening now with data lake houses... Iceberg and some of the newer formats. Which is interesting to me, because there's a ... |
**Justin Garrison:** Appliance. Yeah. |
**Pete Naylor:** Yeah. So it was just meeting with like a hardware partner and looking at what they're building, and thinking about what this thing would look like. They had a lab there, we could go in and see all the blinky lights; that made me happy. It was loud. It was very, very loud, by the way, and hot. They're a... |
**Justin Garrison:** Oh, the bubbles? See it boiling? |
**Pete Naylor:** Two approaches. One where it was pumping water through these pipe channel things and they go to the heat sinks on the CPUs... But there's this other approach where they take the whole piece of equipment and it's submerged... |
**Justin Garrison:** Yeah. You immerse it in like an oil... |
**Pete Naylor:** ...in this fluid in it. |
**Justin Garrison:** Yeah. No heat sinks or anything on it. Just bare chips, and they're just bubbling away. It's amazing. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.