Understanding Your Graph, One Hash at a Time
summary
Tooling that turns federated entity keys into privacy-safe signals by hash entity keys and field/operation/header hashes, stores aggregates in ClickHouse aggregation, and drives cache rules with openfed directives. Demo uses K6 traffic, Redis/Valky L2 plus an L1 per-request map, and shows a rollout with ~50% traffic reaching ~90% served-from-cache and latency dropping from ~51ms to ~28ms. Features include shadow mode, Kafka or mutation-driven invalidation, and recommendations that surface high-cardinality and low-change candidates for safe TTL tuning.
AI-written summary · about
chapters (19)
- 0:00Introduction and welcome
- 1:04Power of entities in graphs
- 3:14Caching challenges in distributed systems
- 5:48Implementing entity caching
- 7:56Four key caching questions
- 9:40Hashing for observability
- 11:29Live demo setup
- 13:15WonderGraph Hub overview
- 15:57Entity intelligence and suggestions
- 17:49Configuring cache rules
- 18:53Testing cache changes
- 20:17Analyzing cache performance
- 22:08Continuous monitoring system
- 25:44Federation and hashing explained
- 26:47L1 vs L2 cache differences
- 30:04Impact of caching on query planning
- 34:09Cache provider support and ownership
- 37:03Client override of cache rules
- 39:09Closing remarks and Q&A
related talks
27mFast DIY O11y with OpenSrc Analytics & Visualization
26mO11y in One: Exploring ClickHouse as a Unified Telemetry Database
35mGraphQL in the Enterprise
26mGraphite@Scale or How to Store Millions of Metrics per Second
18mSwitching Jaeger Distributed Tracing to ClickHouse® to Enable APM
19mFirst Principles of Observability Systems
25mScaling data from the lake to the mesh via OSS
36mThe Graph of Everything - Federated Architecture for Any API Service