Scaling, Caching & Reactivity in Distributed Systems

​Talk 1: Your Cache Isn't as Horizontally Scaled as You Think

​You did everything right. You chose a sharded, in-memory datastore like the excellent Valkey, the Linux Foundation's community-driven fork of Redis. You put a simple, stateless proxy in front to keep the client logic clean. You shipped it. Then, the pager went off.

​This talk is a deep dive into the subtle but catastrophic failure mode where sharded systems fail globally. Drawing on years of experience building and operating some of the internet’s largest distributed systems, Dumanshu, Uber Technical Lead at Airbnb, will dissect how seemingly benign integration patterns can turn a beautifully sharded, OSS Valkey cluster into a system with a single, fragile chokepoint. This is a modern reincarnation of the classic head-of-line blocking problem that plagued HTTP/1.1—a problem many of us thought we had solved. We’ll explore the gnarly interplay of hot shards and client multiplexing that leads to these cascading failures.

​This is not a playbook. Instead, through this case study, you’ll gain:

• ​A deeper intuition for the trade-offs between proxy-based simplicity and the direct-access designs required for microsecond budgets.

• ​The ability to recognize these hidden failure modes and reason about their system-wide impact.

• ​A mental model for achieving true fault isolation, rather than just the illusion of it.

​

​Talk 2: Drasi: React in real-time to changes happening in Distributed Systems

​We will explore how Drasi lets us express sophisticated change detection logic using Graph queries that span data in PostgreSQL, Cosmos DB, Kubernetes, and other sources simultaneously—without writing custom streaming code. Aman, a Principal Engineer in the Azure CTO’s office at Microsoft, will walk through real scenarios (from detecting security vulnerabilities in K8s deployments to building real-time RAG pipelines on structured data for modern LLM-based agents, and optimizing curbside pickup scenarios).

​This will be a technical deep dive with demos, covering both the conceptual framework and practical implementation patterns.