Kafka and Open Draw
Agenda
- 人無法解決的事情 通常程式也沒辦法
- why ? what ? how ?
- why, 為什麼要談它? 我們遇到了什麼困難要解決?
- what, Kafka是什麼?
- how, 如何用它解決我們的困難?
- 開獎設計
why ?
- according ken's last presentation
- hadoop & spark 為我們解決了巨量儲存跟輸出的問題
- 而現在談的是即時的巨量資料處理之於系統的影響
- 突然大量的請求, 即時使用者數據分析, 大量的投注單開獎
what ?
- message queue, first in first out data structure
- just a message queue ?
- multi thread program
- 找到可平行處理的部份, servlet per request per thread, 很多個老闆
- inner queue, nodejs event driven, netty reactor pattern, 分工合作的餐廳
- more feature about queue
- add communication part to be a service
- redis queue
- zeromq
- activemq
- rabbitmq
- kafka (zookeeper for cluster)
- profile of kafka
- special part of kafka
- persist first, leveldb
- push, pull, mix and consumer offset
- consumer group
- multi partition
- isr and replication
- almost once, at least once, exactly-once semantics
- batch and stream process (spark, storm, kafka)
- stream aggregate 拓墣結構
- connect
how ?
- 開獎設計
- real case performance
http://spring.io/projects/spring-cloud-zookeeper