TL;DR
Get the latest gadgets delivered free — and shop member deals
- Fast, free delivery on millions of items
- Access to Prime Big Deal Days deals on October 6–7
- Prime Video, Amazon Music and more included
In an episode of The Pragmatic Engineer, Cockroach Labs co-founder Peter Mattis discusses lessons from building storage systems at Google and distributed databases at Cockroach Labs. He also describes how AI has brought him back to writing code and argues it can increase the impact of experienced engineers; the episode does not provide independent measurements of those productivity claims.
The Pragmatic Engineer has published an interview with Peter Mattis, co-founder and chief technology officer of Cockroach Labs, covering how engineers build fast, reliable distributed storage and how AI is affecting his own coding work. Mattis, who previously worked on Gmail and distributed storage at Google, describes technical tradeoffs from those systems and says AI has helped him return to writing code after his work shifted toward management.
Mattis discusses storage decisions behind early Gmail, including the use of B-trees to track email threads and unread counts. The source says incoming messages were matched to threads through the search index, with B-trees keeping track of threads and their unread counts. The interview places this work within the broader challenge of making storage systems operate correctly as they grow.
At Google, Mattis and colleagues also worked on Colossus, the successor to the Google File System. According to the source, the earlier system kept three full copies of stored data. Colossus used Reed–Solomon erasure coding to store data twice while increasing redundancy, and reduced storage overhead by 33%. The report does not provide a separate technical paper or methodology for that figure.
Mattis also recounts building alternatives to standard library data structures after profiling showed performance costs. He says a B-tree implementation replaced uses of C++’s std::map in Google software, improving speed through spatial locality while using fewer pointers. He later built a faster Swiss Table implementation for Go; the Go team helped complete work that eventually entered the language’s library, according to the report.
Storage Choices Shape System Reliability
The examples show how database performance depends on data layout, redundancy and network delay, as well as on the choice of a database product. For Gmail, tracking messages within threads required structures that could support lookup and updates. For Colossus, the storage team changed how data was encoded, aiming to reduce the amount stored while maintaining redundancy. Those choices affect the cost and reliability of services that hold large volumes of user data.
Mattis’s examples of replacing library structures also underline that familiar tools can become bottlenecks under particular workloads. His account is a practitioner’s perspective, rather than a controlled comparison across systems. The interview connects these engineering lessons to AI: Mattis says it lets experienced engineers spend more time producing code, but the source does not quantify the effect or establish that all teams will see the same results.
Top picks for "distribut databas peter"
As an affiliate, we earn on qualifying purchases.
From GIMP to Google Storage
Mattis first became known as an original creator of GIMP, the open-source image editor he developed with college roommate Spencer Kimball. The source says the first version of Google’s logo was made with GIMP. It also recounts that Mattis initially declined an offer from Sergey Brin because of the commute from San Francisco to Mountain View, then joined Google after the company contacted him again.
At Google, Mattis worked on Gmail and distributed storage before co-founding Cockroach Labs, which develops the distributed SQL database CockroachDB. The interview returns repeatedly to B-trees, linking their use in Gmail’s storage layer, an alternative to std::map, and CockroachDB’s range index. The source points to the paper “The Ubiquitous B-Tree” as further discussion of their recurring role in storage systems.
The episode also compares network timing across eras. The source says a within-zone network round trip that took milliseconds when Colossus was built takes about 100 microseconds today, described there as 10 times faster. It also recounts Mattis’s view that long-distance latency can be constrained by the speed of light, including the possibility of routes through space. These are examples from the conversation, not a general guarantee about current network performance.
“There’s always going to be someone else working on your idea.”
— Peter Mattis, as quoted in The Pragmatic Engineer
AI Productivity Claims Need Measures
The source describes Mattis’s view that AI has made him more productive without lowering quality, and that it has brought him back to coding after a period focused more on management. It does not give measured productivity results, define how quality was assessed or compare AI-assisted output with a baseline. The reported experience should be read as his account, not as evidence that the same gains apply across engineering teams.
The source material also does not specify the episode’s publication date, include a full transcript, or provide independent documentation for the numerical claims about Colossus and network latency. The details of how Mattis’s current AI workflow operates, and how code review may change, remain open in the material provided.
Follow the Episode and Storage Work
The Pragmatic Engineer says the episode is available on YouTube, Apple and Spotify, with a transcript and timestamps on its page. Listeners can hear Mattis discuss distributed database engineering, the history behind his work and his views on AI-assisted coding. The source gives no upcoming product launch or research milestone tied to the interview.
Separately, the report points readers to turbopuffer’s published documentation of its storage architecture rewrite at turbopuffer.com/v3. That project is presented as related reading on storage design; the source does not say it is part of Mattis’s work or specify when its next update will appear.
Key Questions
Who is Peter Mattis?
Peter Mattis is co-founder and CTO of Cockroach Labs. The source also identifies him as an original creator of GIMP and a former Google engineer who worked on Gmail and distributed storage.
What did the interview say about Colossus?
According to the source, Colossus succeeded the Google File System and used Reed–Solomon erasure coding. It says the system stored data twice while increasing redundancy and cut storage overhead by 33%.
What role did B-trees play in the examples?
The report says B-trees were used to track Gmail threads and unread counts. Mattis also describes building a B-tree alternative to C++’s std::map, and the interview notes B-trees’ use in CockroachDB’s range index.
Does the episode prove AI makes engineers more productive?
No. Mattis describes his own experience of increased productivity and maintained quality, but the source provides no measured results or comparison baseline to establish a general effect.
Source: rss
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
