Podcast Archives

The episode contains substantive ideas about database evolution, particularly around AI's impact on data queries, the shift from exact to approximate results, and specific technical solutions like parameterized secure views and filtered vector search. However, much time is spent on foundational concepts and product announcements rather than densely packed novel insights. Several minutes involve basic product descriptions and repetitive explanations.

“What we are seeing is the confluence of structured and unstructured data. And what that's leading is applications, workloads, where people are expecting that these things get put together. And now the database world starts to look a little bit more like the search world, or in academic terms it looks like information retrieval.”

“Without it there's no way to get the kind of security you need for agentic applications.”

The discussion offers some fresh framing - particularly around parameterized secure views as a foundational security primitive and the idea that determinism itself must be abandoned in the AI era. However, many concepts (vector search in databases, multi-modal data, schema governance for AI) are increasingly common in the current discourse. The core themes around Spanner's global distribution and consistency have been well-established in previous work.

“Parameterized secure views take us to the next level where you can apply a parameter...the kind of parameter that you might apply in the where clause here is essentially the same check for Authentication the authorization authorized access which I mentioned is there in three tier web applications.”

“I think we have to embrace the chaos. And that means putting ourselves in this kind of weird spot...I think the fundamentally big thing that's changed for my area databases is that we're dealing with structured and unstructured data together.”

Silesh Krishnamurthy is a highly credible guest with 25+ years in databases across IBM, startup (acquired by Cisco), AWS Aurora, and now VP of Engineering for Databases at Google. He has directly built or led some of the most significant database infrastructure in the world and currently manages both Google Cloud's transactional databases and all of Alphabet's operational databases. This is a rare combination of seniority, hands-on experience, and current relevance.

“I have been in the database industry for most of my professional career. Started in the mid-90s. Um, I worked at IBM, worked on DB2...went to grad school...did my PhD on streaming databases...started a company around to try to commercialize my PhD work...sold our company to Cisco...went to Amazon, I worked at AWS and I worked on a system called Aurora...came to Google in 2019”

“I lead transactional databases for Google Cloud, but I have a second job. I also lead operational databases for all of Alphabet. And so whether it's Gemini or Gmail or YouTube, all of these run that same database, uh, infrastructure on the same set of systems and Spanner and BigTable.”

The episode includes specific technical details (HNSW vs SCAN vector algorithms, 6x faster queries, 4x less memory usage, Spanner Omni single binary packaging, atomic clocks for TrueTime, parameterized secure views). However, many claims lack concrete customer examples, quantified impact metrics, or timelines. Product announcements dominate over detailed case studies. The DanaPay and SoundCloud examples are mentioned but not deeply analyzed.

“In our tests we can get 6x faster vector queries than HNSW, PGvector and standard Postgres. We use 4x less memory.”

“They're using Spanner Graph for their anti money laundering use case. And by running these algorithms directly within Spanner, they can do a whole bunch of things. They can detect suspicious networks much faster again without having to go through an ETL cycle.”

The host asks solid foundational questions and shows genuine curiosity (e.g., about filtered vector search, GenUI, Spanner Omni architecture). However, follow-ups are often soft and don't push back on claims. The host allows tangential discussions without redirecting, and rarely challenges assumptions. Questions tend toward 'tell me more' rather than 'does this really work' or exploring contradictions.

“So for those of my fellow Spanner noobs. So why would I reach for something like a Spanner instead of a Postgres, my favorite. Or MySQL or something like that?”

“So basically what I'm hearing is, so, especially with TrueTime, the level of accuracy needed to scale. At Google levels, you need that level of accuracy.”