How do vector (search) databases work? ft: turbopuffer
For memberships: join this channel as a member here:https://www.youtube.com/channel/UC_mGuY4g0mggeUGM6V1osdA/joinSummary:In this conversation, Kaivalya Apte and Simon Eskildsen talk about vector databases, particularly focusing on TurboPuffer. They discuss the importance of vector search, embeddings, and the challenges associated with building efficient search engines. The conversation covers various aspects such as cost considerations, chunking strategies, multi-tenancy, and performance optimization. Simon shares insights on the future of vector search and the significance of observability and metrics in database performance. The discussion emphasizes the need for practical application and experimentation in understanding these technologies.Chapters:00:00 Introduction to Vector Databases10:34 Understanding Vectors and Embeddings15:03 Example: Designing a Search Engine for Podcasts27:53 Scaling Challenges in Vector Search36:46 Indexing and Querying in TurboPuffer38:12 Understanding Indexing and Query Planning45:45 Exploring Index Types and Their Performance50:27 Data Ingestion and Embedding Retrieval54:19 Use Cases and Challenges in Vector Search01:01:22 Metrics and Observability in Vector Databases01:03:52 Future Trends in Vector Search and DatabasesReferences:How do build a database on Object Storage? https://youtu.be/RFmajOeUKnETurbopuffer https://turbopuffer.com/Continous Recall measurement: https://turbopuffer.com/blog/continuous-recallTurbopuffer architecture: https://turbopuffer.com/architecture
--------
1:08:58
Are your Data Pipelines Complex?
The GeekNarrator memberships can be joined here: https://www.youtube.com/channel/UC_mGuY4g0mggeUGM6V1osdA/joinMembership will get you access to member only videos, exclusive notes and monthly 1:1 with me. Here you can see all the member only videos: https://www.youtube.com/playlist?list=UUMO_mGuY4g0mggeUGM6V1osdA------------------------------------------------------------------------------------------------------------------------------------------------------------------About this episode: ------------------------------------------------------------------------------------------------------------------------------------------------------------------In this conversation, Jacopo and Ciro discuss their journey in building Bauplan, a platform designed to simplify data management and enhance developer experience. They explore the challenges faced in data bottlenecks, the integration of development and production environments, and the unique approach of Bauplan using serverless functions and Git-like versioning for data. The discussion also touches on scalability, handling large data workloads, and the critical aspects of reproducibility and compliance in data management. Chapters:00:00 Introduction03:00 The Data Bottleneck: Challenges in Data Management06:14 Bridging Development and Production: The Need for Integration09:06 Serverless Functions and Git for Data17:03 Developer Experience: Reducing Complexity in Data Management19:45 The Role of Functions in Data Pipelines: A New Paradigm23:40 Building Robust Data Solutions: Versioning and Parameters30:13 Optimizing Data Processing: Bauplan Runtime46:46 Understanding Control Planes and Data Management48:51 Ensuring Robustness in Data Pipelines52:38 Data Quality and Testing Mechanisms54:43 Branching and Collaboration in Data Development57:09 Scalability and Resource Management in Data Functions01:01:13 Handling Large Data Workloads and Use Cases01:09:05 Reproducibility and Compliance in Data Management01:16:46 Future Directions in Data Engineering and Use CasesLinks and References:Bauplan website:https://www.bauplanlabs.com
--------
1:23:28
Can Math simplify incremental compute?
In this episode of The Geek Narrator podcast, Lalit Suresh, CEO of Feldera, joins us to share insights on incremental view maintenance and its significance in modern data processing.We have discussed the challenges posed by distributed systems, the mathematical foundation of DBSP, and how Feldera's architecture addresses these challenges. Performance optimization, handling late events, and the future of stream processing, the importance of SQL in creating efficient data workflows - its all in here.Chapters00:00 Introduction to Incremental View Maintenance06:30 Challenges in Distributed Systems11:46 Batch Processing vs Stream Processing16:27 Understanding DBSP: The Mathematical Foundation27:46 Architecture of Feldera and Data Flow39:23 Partitioning and Storage Layer in Feldera42:51 Understanding Co-Design Storage Layers45:52 Foreground and Background Workers in DBSP49:16 Tuning Background Workers for Performance49:41 Synchronous Compute Model and View Propagation51:35 Zsets and Batch Processing in Stream Workloads54:00 Data Model Optimization in Feldera57:22 Handling Late Events and Lateness in Feldera01:01:18 Watermarks and Lateness Annotations01:04:20 Error Handling and Idempotency in Feldera01:11:05 Feldera's Differentiators and Future Roadmap
--------
1:17:13
Redpanda - High Performance Streaming Platform for Data Intensive Applications
The GeekNarrator memberships can be joined here: https://www.youtube.com/channel/UC_mGuY4g0mggeUGM6V1osdA/joinMembership will get you access to member only videos, exclusive notes and monthly 1:1 with me. Here you can see all the member only videos: https://www.youtube.com/playlist?list=UUMO_mGuY4g0mggeUGM6V1osdA------------------------------------------------------------------------------------------------------------------------------------------------------------------About this episode: ------------------------------------------------------------------------------------------------------------------------------------------------------------------In this conversation, Alex from Red Panda discusses his engineering background, the challenges faced in reliability engineering, and the journey of building a better streaming system. He emphasizes the importance of understanding latency and performance in engineering systems, the market position of Red Panda in relation to Kafka, and the complexities involved in optimizing codebases for better performance. In this conversation, Alex discusses Red Panda's architecture, focusing on its thread architecture, memory allocation mechanics, and the importance of protocol correctness. He highlights how Red Panda stands out in the data systems landscape by eliminating unnecessary complexities and optimizing performance across various latency spectrums. The discussion also touches on the future of data processing, emphasizing the shift towards agentic workloads and the integration of analytical and operational layers.Chapters00:00 Introduction11:07 Building a Better Streaming System19:10 Market Position and Competition25:06 Optimizing Latency and Performance32:38 Understanding Complexity in Codebases33:36 Thread Architecture and Concurrency Models39:39 Memory Allocation Mechanics47:31 Protocol Correctness and Optimization Strategies56:27 Red Panda's Unique Position in Data Systems01:02:05 The Future of Data Processing and Agentic WorkloadsBlogs:TPC buffers: https://www.redpanda.com/blog/tpc-buffershttps://www.redpanda.com/blog/always-on-production-memory-profiling-seastarhttps://www.redpanda.com/blog/end-to-end-data-pipelines-types-benefits-and-process------------------------------------------------------------------------------------------------------------------------------------------------------------------Like building real stuff?------------------------------------------------------------------------------------------------------------------------------------------------------------------Try out CodeCrafters and build amazing real world systems like Redis, Kafka, Sqlite. Use the link below to signup and get 40% off on paid subscription.https://app.codecrafters.io/join?via=geeknarrator------------------------------------------------------------------------------------------------------------------------------------------------------------------Link to other playlists. LIKE, SHARE and SUBSCRIBE------------------------------------------------------------------------------------------------------------------------------------------------------------------If you like this episode, please hit the like button and share it with your network. Also please subscribe if you haven't yet.Database internals series: https://youtu.be/yV_Zp0Mi3xsPopular playlists:Realtime streaming systems: https://www.youtube.com/playlist?list=PLL7QpTxsA4se-mAKKoVOs3VcaP71X_LA-Software Engineering: https://www.youtube.com/playlist?list=PLL7QpTxsA4sf6By03bot5BhKoMgxDUU17Distributed systems and databases: https://www.youtube.com/playlist?list=PLL7QpTxsA4sfLDUnjBJXJGFhhz94jDd_dModern databases: https://www.youtube.com/playlist?list=PLL7QpTxsA4scSeZAsCUXijtnfW5ARlrsNStay Curios! Keep Learning!#streaming #kafka #redpanda #c++ #databasesystems #SQL #distributedsystems #memoryallocation #garbagecollection
--------
1:05:28
Hosted PostgreSQL on bare metal and uni kernel
The GeekNarrator memberships can be joined here: https://www.youtube.com/channel/UC_mGuY4g0mggeUGM6V1osdA/joinMembership will get you access to member only videos, exclusive notes and monthly 1:1 with me. Here you can see all the member only videos: https://www.youtube.com/playlist?list=UUMO_mGuY4g0mggeUGM6V1osdA------------------------------------------------------------------------------------------------------------------------------------------------------------------About this episode: ------------------------------------------------------------------------------------------------------------------------------------------------------------------In this episode, we talk to Søren Schmidt, Co-Founder and CEO of Prisma, discussing the evolution of Prisma from a backend as a service to a popular ORM and now to Prisma Postgres. He shares insights into the challenges faced during this journey, the importance of user feedback, and the innovative architecture of Prisma Postgres, which leverages micro VMs for performance optimization. The conversation also touches on the complexities of managing data centers and the strategies employed to ensure a seamless user experience. In this conversation, Søren Schmidt discusses the details about Postgres snapshots, their impact on performance, and the mechanisms for fault tolerance. He explains how Pulse change data capture works and how Prisma Postgres simplifies database management for users. Chapters00:00 Introduction to Prisma and Its Evolution03:00 The Journey from ORM to Prisma Postgres06:00 Simplifying Database Management09:01 Understanding Prisma Postgres Architecture12:12 The Role of Accelerate in Query Routing14:51 Optimizing Query Processing with Micro VMs18:12 Maintaining Postgres Integrity in a Micro VM Environment21:07 User Experience and Community Feedback23:57 Challenges of Data Center Management27:09 Cold Starts and Performance Optimization34:30 Understanding Snapshots in Postgres38:55 Snapshot Mechanisms and Fault Tolerance44:09 Change Data Capture with Pulse55:07 Transitioning to Prisma Postgres58:45 Community and Getting Started with Prisma PostgresSome blogs worth checking out:https://www.prisma.io/blog/prisma-postgres-the-future-of-serverless-databaseshttps://www.prisma.io/blog/cloudflare-unikernels-and-bare-metal-life-of-a-prisma-postgres-queryhttps://www.prisma.io/blog/announcing-prisma-postgres-early-accessPrisma Postgres relies heavily on the Unikraft project. There is a good introductory talk here: https://www.youtube.com/watch?v=n4wOyAuNhl0And some very technical papers here: https://unikraft.org/community/papersThe best way to get started with Prisma Postgres is to go straight to https://www.prisma.io/ ------------------------------------------------------------------------------------------------------------------------------------------------------------------Like building real stuff?------------------------------------------------------------------------------------------------------------------------------------------------------------------Try out CodeCrafters and build amazing real world systems like Redis, Kafka, Sqlite. Use the link below to signup and get 40% off on paid subscription.https://app.codecrafters.io/join?via=geeknarrator------------Database internals series: https://youtu.be/yV_Zp0Mi3xsPopular playlists:Realtime streaming systems: https://www.youtube.com/playlist?list=PLL7QpTxsA4se-mAKKoVOs3VcaP71X_LA-Software Engineering: https://www.youtube.com/playlist?list=PLL7QpTxsA4sf6By03bot5BhKoMgxDUU17Distributed systems and databases: https://www.youtube.com/playlist?list=PLL7QpTxsA4sfLDUnjBJXJGFhhz94jDd_dModern databases: https://www.youtube.com/playlist?list=PLL7QpTxsA4scSeZAsCUXijtnfW5ARlrsN
The GeekNarrator podcast is a show hosted by Kaivalya Apte who is a Software Engineer and loves to talk about Technology, Technical Interviews, Self Improvement, Best Practices and Hustle.
Connect with Kaivalya Apte https://www.linkedin.com/in/kaivalya-apte-2217221a
Tech blogs: https://kaivalya-apte.medium.com/
Wanna talk? Book a slot here: https://calendly.com/speakwithkv/hey
Enjoy the show and please follow to get more updates. Also please don’t forget to rate and review the show.
Cheers