Database
Replication, sharding, caching, and change-data-capture topologies.
Primary with read replicas
Writes to one, reads from many. Two regions carve out the "write side" and "read side" of the data tier; the primary is sizeScale-emphasised.
doc { cols: 4, rows: 2 }
region @B1:B2 "Write path" color=accent/30
region @C1:D2 "Read-only" color=primary/28
icon :app @A1 tabler/server "App"
icon :pri @B1 tabler/database "Primary" sizeScale=1.4
icon :r1 @C1 tabler/database-export "r-1"
icon :r2 @D1 tabler/database-export "r-2"
app --> pri "writes" width=2
app --> r1 "reads"
app --> r2 "reads"
pri --> r1 dash="2 4"
pri --> r2 dash="2 4"
note @C2 (r1, r2) "Async\nlag ≤ 5s"Sharding by key
Plain layout — three shards, one router. A deliberate contrast with the richer example above.
doc { cols: 3, rows: 3 }
icon :app @A2 tabler/server "App"
icon :router @B2 tabler/arrows-shuffle "Router" sizeScale=1.3
icon :s1 @C1 tabler/database "Shard A"
icon :s2 @C2 tabler/database "Shard B"
icon :s3 @C3 tabler/database "Shard C"
app --> router
router --> s1
router --> s2
router --> s3Read-through cache
Two frames contrast the hot and cold paths. Hover and flip between them — frame 1 is the fast round-trip served from cache; frame 2 is the miss falling back to the DB with a "populate on read" note.
Frame 1 / 1–2
CDC → downstream
Four downstream consumers fed off the primary. Badges mark which feeds are healthy vs catching up.
doc { cols: 4, rows: 2 }
icon :db @A1 tabler/database "Primary" sizeScale=1.3
icon :cdc @B1 tabler/arrows-shuffle "CDC"
icon :search @C1 tabler/search "Search" { badges: ['check'] }
icon :bi @D1 tabler/chart-dots "BI" { badges: ['check'] }
icon :cache @C2 tabler/bolt "Cache" { badges: ['alert'] }
icon :ml @D2 tabler/brain "ML" { badges: ['check'] }
db --> cdc
cdc --> search
cdc --> bi
cdc --> cache dash="2 4"
cdc --> ml