Catalogue
- Cards
- 165
- Records
- 486
- Workloads
- 19
- Corpus version
- 74847648e25a
20 of 486 records match.
| Kind | Record | Organisation | Year | Setup |
|---|---|---|---|---|
| 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 | Batch ping writes with COPY in one transaction instead of calling a stored procedure once per ping | Healthchecks.io | 2025 | One database |
| Decision | Cache not-found results for UUID ping URLs so repeated pings to missing checks skip the database | Healthchecks.io | 2025 | One database |
| Decision | Scale ZippyDB keyspaces online via logical-to-physical shard indirection and coordinator-worker resharding | Meta | 2023 | Sharded |
| Decision | Offload Instagram-to-Threads follow-graph mirroring and notifications to rate-limited Async workers | Meta | 2023 | Sharded |
| Failure | A viral customer's pageviews backlog the shared queue for all sites while the database runs out of disk IOPS | Fathom Analytics | 2020 | One database |
| Failure | Unindexed backlog tables and a text-typed site ID kept the database slow after the disk was fixed | Fathom Analytics | 2020 | One database |
| Decision | Cap concurrent queue workers on purpose to protect the database, accepting a bigger backlog | Fathom Analytics | 2020 | One database |
| Decision | Limit pageviews processed per site per minute so one viral site cannot block the rest, then drop it | Fathom Analytics | 2020 | One database |
| Failure | Database contention filled Redis memory, and a full Redis could neither accept nor release jobs | Slack | 2017 | Sharded |
| Decision | Added Kafka in front of the Redis job queue instead of replacing Redis | Slack | 2017 | Sharded |
| Decision | Enqueue path acknowledges after the Kafka leader only, trading a small risk of lost jobs for latency | Slack | 2017 | Sharded |
| Decision | Serve every Q&A site from one multi-tenant application and keep web servers overprovisioned | Stack Overflow | 2016 | Sharded |
| Decision | Use Elasticsearch for search instead of SQL Server full-text search, because SQL CPUs are expensive | Stack Overflow | 2016 | Sharded |
| Decision | Cache in two layers, a web-server HTTP cache and Redis, and clear the first layer through Redis pub/sub | Stack Overflow | 2016 | Sharded |
| Decision | Push updates to browsers over websockets, which are more efficient than polling, at the cost of load balancer port exhaustion | Stack Overflow | 2016 | Sharded |
| Failure | A hundred checker Lambdas each opened a database connection at once, and Postgres refused further connections | OnlineOrNot | One database | |
| Decision | Fan results in through SNS and a batching queue so one Lambda writes to Postgres, instead of upgrading the database | OnlineOrNot | One database |