Apache Cassandra Meetup

We welcome you to join us for an in-person meetup hosted by Uber and the Apache Cassandra community. We will kick off with opening remarks by Osama Mazahir, Director of Storage Platform at Uber followed by a series of interesting talks by the speakers from Uber, Netflix, Apple, and DataStax.

Event Details

• This Meetup is an in-person event only
• Registration is required for the Meetup. Please RSVP & answer the questions (full name & email address)
• Event location details will be emailed a few days before the event to those who have registered for the event and provided an email address (not located at pin location)

Event Agenda

• 5:30 PM - Networking and Snacks
• 6:30 PM - Opening Remarks
• 6:40 PM - Talk 1: Auto Repair in Apache Cassandra: CEP-37 Scheduler in Action
• 6:55 PM - Talk 2: Cassandra Ecosystem Roadmap: Multi-Model Architecture and Disaggregation Through Ecosystem Integration
• 7:10 PM - Talk 3: Resilient Cassandra: Intelligent Backpressure Using Resource Utilization and Queueing Metrics
• 7:25 PM - Talk 4: Contributing to Apache Cassandra: An easy guide to getting started
• 7:40 PM - Talk 5: Five things about Cassandra 5 you didn’t know
• 7:55 PM - Talk 6: Cassandra Client Standardization at Uber: Service Discovery, Observability, and Enforcement
• 8:10pm: Closing remarks

Presentation Info

• Talk 1: Auto Repair in Apache Cassandra: CEP-37 Scheduler in Action
  Repair is critical to maintaining consistency in Apache Cassandra, yet has historically been operationally expensive and fragile—especially at scale. This talk introduces Auto Repair, a native, intelligent repair scheduler contributed through CEP-37 and now integrated into Cassandra’s trunk and Uber’s production environments. Jaydeepkumar and Runtian will cover why repair matters, the operational pain points of traditional approaches, and how the Auto Repair Scheduler eliminates manual toil by running repairs seamlessly in the background—just like compaction. With features like dynamic configuration, intelligent token-range splitting, disk-aware guardrails, and rich observability, Auto Repair transforms repair into a self-managing, production-grade capability. Whether you're operating a handful of nodes or managing thousands, this session will show how Auto Repair makes Cassandra more resilient, scalable, and operator-friendly—without external tooling or custom scripts.
  - Speaker Bios:
  Jaydeepkumar Chovatia is a Senior Staff Engineer at Uber and a committer to the Apache Cassandra project. He designs scalable, fault-tolerant storage systems and leads key initiatives across Uber's storage infrastructure. Jaydeep is known for his architectural leadership in open-source and internet-scale systems, with contributions that have shaped the reliability and resilience of platforms used across the industry.
  - Runtian Liu is a Staff Software Engineer at Uber, where he is a core member of the Storage Cassandra team. He specializes in building and scaling large-scale distributed storage systems that power Uber's critical infrastructure. Runtian has led the company's Cassandra modernization efforts, driving initiatives that significantly improved the stability and operational efficiency of Uber's Cassandra fleet. His work includes the design and implementation of Cassandra Auto-Repair 2.0 and a major enhancement to the system's graceful replacement process—both of which have become foundational to Uber's Cassandra operations. Most recently, Runtian has been focused on making materialized views production-ready in Cassandra, contributing to the broader goal of evolving Cassandra's feature set for real-world, high-throughput workloads.
• Talk 2: Cassandra Ecosystem Roadmap: Multi-Model Architecture and Disaggregation Through Ecosystem Integration
  Let's talk about a potential future extending Cassandra beyond an isolated distributed database into a multi-model system with integrated search (via CDC to Solr/OpenSearch) and analytics capabilities. Changes to the cassandra-sidecar could enable automated cluster scaling, declarative repair and compaction management, and native object storage backup/restore. Key architectural changes could include: proactive tombstone management with threshold-triggered repairs and compaction, disaggregated analytics operations, and direct data lake integration via bulk snapshot and iceberg export. We'll discuss how cassandra-analytics could handle LSM-tree maintenance out-of-band and cover the value of bringing a 1st class kubernetes operator in-tree. This talk covers the technical design decisions and implementation challenges we can expect to face as we move into a more integrated ecosystem future.
  - Speaker Bios:
  Josh McKenzie from Apple
  Josh has been writing code since he first got a pair of TRS-80s decades ago. He has worked on distributed systems for nearly 20 years, from High Frequency Trading platforms to Apache Cassandra since 2014. As a PMC member and committer on Apache Cassandra, he authored Change Data Capture and previously served as VP of Engineering and head of Open Source Strategy at DataStax. He now works at Apple as part of the Cassandra team where he's spearheaded upstreaming a backlog of work to open source and is driving cross-project collaboration between many users of and contributors to the database. His current focus is on Cassandra's architectural integration with big data analytics, indexing, and large-scale data lakes.
  - Doug Rohrer from Apple
  Doug's been a developer for over 30 years. He's worked in government and industry in areas ranging from occupational safety and health to finance to a home shopping network. His interest in distributed systems was sparked about 10 years ago when he started writing Elixer and Erlang, which took him to Basho where he worked on Riak for several years. He now works at Apple as part of the Cassandra team, where he's contributed to both the core Cassandra database and the Analytics library.
• Talk 3: Resilient Cassandra: Intelligent Backpressure Using Resource Utilization and Queueing Metrics
  In distributed systems, safeguarding downstream datastores like Cassandra from overload is critical for maintaining availability and performance. Traditional backpressure mechanisms often rely on simplistic metrics such as CPU utilization or raw latency, which can lead to misguided load shedding and violate the "Do No Harm" principle. This talk introduces a sophisticated, metric-driven backpressure system that leverages Cassandra's internal resource utilization metrics—specifically PSI (Pressure Stall Information) queueing data, normalized by QPS—to accurately attribute resource strain and trigger intelligent load shedding. We'll explore the limitations of conventional metrics, the rationale behind using queueing-based indicators, and how normalizing by request rate disambiguates true overload from benign latency spikes. Attendees will gain practical insights into building resilient, self-protecting data tiers that optimize both availability and performance, even under fluctuating workloads and query complexities.

• Speaker Bios:
  Vidhya Arvind is a Tech Lead at Netflix and a founding architect of its cutting-edge data abstraction platform. She specializes in designing robust APIs and high-performance abstractions that power seamless data access at massive scale. Known for her systems thinking and curiosity, Vidhya thrives on debugging, innovating, and solving deeply technical challenges. Her work has been instrumental in evolving Netflix's data infrastructure, particularly in building mission-critical systems on Cassandra that deliver exceptional efficiency, reliability, and resilience.
  - William Schor is a Senior Software Engineer at Netflix, where he works on the KeyValue team. He focuses on building and scaling distributed systems, with a particular interest in resilience and high availability. At Netflix, William plays a key role in hardening and evolving the KeyValue platform—which includes managing the Cassandra fleet and abstraction layer—to support a wide range of fast-growing and diverse use cases across the company.

• Talk 4: Contributing to Apache Cassandra: An easy guide to getting started
  Want to contribute to Apache Cassandra but don't know where to start? You're not alone! In this talk we will explore various dimensions of contributing to the Apache Cassandra community. You will learn about the various roles in an ASF Project, types of contributions, the Apache Cassandra project's governance structure and key tools and resources to get you started in your journey to contribute to the project. We will finally talk about the progression from a contributor to the PMC role in the project.

• Speaker Bio:
  Dinesh Joshi from Apple
  Dinesh is the PMC Chair / VP of the Apache Cassandra project. He is also an Engineering Manager at Apple leading Cassandra. For the past two decades Dinesh has played various roles including that of a Principle Software Engineer working on highly scalable distributed systems and databases.

• Talk 5: Five things about Cassandra 5 you didn’t know
  What if I told you that all those carefully crafted denormalized tables you've been maintaining might soon be obsolete? Or that Cassandra just quietly became an AI database while you weren't looking? In this talk, we'll uncover five surprising capabilities in Cassandra 5.0 that will make you question everything you thought you knew about distributed data modeling. While you were happily running Cassandra 3.0 we went and changed everything! Warning: You will want to upgrade immediately after seeing this.

• Speaker Bio:
  Patrick McFadin is the Principal Technical Strategist at DataStax. He is also a committer and PMC member on the Apache Cassandra project and a co-author of the O’Reilly book “Managing Cloud-Native Data on Kubernetes.” Before joining DataStax, he held positions as Chief Architect, Engineering Lead, and Database DBA/Developer.

• Talk 6: Cassandra Client Standardization at Uber: Service Discovery, Observability, and Enforcement
  At a large scale, it is very easy to have diverse integration patterns on the client side, which may bring unpleasant surprises. In this talk, Java and Long will take us through a journey showing how Uber standardizes the client side configurations, policy enforcements, observability and service discovery mechanism.
  - Speaker Bios:
  Java Servin - Staff Software Engineer from Uber
  Extensive expertise in distributed data architectures and micro services. He is a senior member in the Cassandra team, with a focus on building and optimizing high-performance, resilient data solutions. He leads critical efforts to standardize and modernize the Cassandra client, significantly enhancing performance and reliability for an ecosystem of over a thousand services.
  - Long Pan Sr. Software Engineer from Uber
  Member of the Cassandra team at Uber. He is deeply involved in a broad range of activities to maintain and enhance Cassandra at scale as a distributed database. His work spans operational improvements and development projects. Outside of work, he enjoys cooking and exploring new places.