Catalogue
- Cards
- 165
- Records
- 486
- Workloads
- 19
- Corpus version
- 74847648e25a
27 of 486 records match; the first 20 are shown.
| Kind | Record | Organisation | Year | Setup |
|---|---|---|---|---|
| Failure | Traffic peak overloaded one datacenter's load balancers while a service-mesh sidecar failed to scale, and a client retry bug amplified authentication traffic | GitHub | 2026 | Multi-region |
| Decision | Replace managed object storage with Versity S3 Gateway on one server, backed up by rsync, to stop provider slowdowns | Healthchecks.io | 2026 | One server |
| Failure | Connection pools that recycled every 15 minutes synchronized into a thundering herd after a managed Postgres upgrade | Clerk | 2025 | |
| Decision | Use cloud-based AWS Elemental transcoding paired with a custom packager and custom Live Origin | Netflix | 2025 | Multi-region |
| Decision | Deliver live video over the existing Open Connect CDN shared with on-demand streaming | Netflix | 2025 | Multi-region |
| Decision | Stream live video over HTTPS with 2-second segments and cloud-delivered segment-template manifests | Netflix | 2025 | Multi-region |
| Decision | Validate live changes with internal test streams, synthetic load up to 100K starts/sec, failure injection, and scheduled live practice | Netflix | 2025 | Multi-region |
| Decision | Pair viewership-prediction pre-provisioning with prioritized load shedding and feature degradation | Netflix | 2025 | Multi-region |
| Failure | Recurring brief live viewing interruptions triggered manual stream restarts and 10x traffic increases | Netflix | 2025 | Multi-region |
| Failure | Users who follow hundreds of thousands of accounts made hot shards that slowed the timelines sharing them | Bluesky | 2025 | Sharded |
| Decision | Drop timeline writes with a probability set by how many accounts a user follows, accepting imperfect timelines | Bluesky | 2025 | Sharded |
| Decision | Cache high-follow accounts in a Redis sorted set loaded into memory every 30 seconds instead of reading follow counts per write | Bluesky | 2025 | Sharded |
| Decision | Replace the monolithic Reloaded video pipeline with microservices on Cosmos and switch over by use case | Netflix | 2023 | |
| Decision | Split video stages by domain boundary while hiding chunked encoding and assembly inside VES | Netflix | 2023 | |
| Decision | Separate member streaming and studio operations into dedicated workflow orchestrators over shared video services | Netflix | 2023 | |
| Decision | Acknowledge archives only after they reach standard S3, not the reduced-redundancy class | Tarsnap | 2023 | |
| Decision | Run the server on one EC2 instance and accept an outage if its hardware fails | Tarsnap | 2023 | |
| Failure | An EC2 power loss corrupted the metadata disk, and a full rebuild from S3 ran slowly through recovery code only ever tested on small data | Tarsnap | 2012 | |
| Decision | Keep all user data in S3 and only cached metadata on the EC2 server so that all server state can be rebuilt | Tarsnap | 2012 | |
| Decision | Treat the corrupted local state as untrustworthy and rebuild it all from S3 instead of trusting it | Tarsnap | 2012 |
7 more records. Sign in to see all 27.