movieflix
image.png- topic:
video play position:
a topic that can have multiple producers;
should be highly distributed if high volume > 30 partitions;
choose "user_id" as a key, to make sure all my user's data in order;
recommendations
low volume topic
kafka streams recommendation source data from analytical training
GetTaxi
requirement
image.pngsocial media architecture of kafka
basically decoupled different components, we can call the model CQRS(command query responsibility segregated)
image.png- format data as an event.
User_234 liked post_123 at 3 am
image.png- big data ingestion
real time: like spark/storm
batch: hadoop/RBMS
monitoring
image.pngimportant metrics
two types of index:
image.pngimage.png
advertised hostname and listeners:
image.png
网友评论