一：分佈式緩存redis 單進程單線程（一個redis進程是一個redis實例）
https://blog.csdn.net/liupeng_qwert/article/details/77263187
https://www.cnblogs.com/syyong/p/6231326.html
https://help.aliyun.com/document_detail/52228.html?spm=a2c4g.11186623.6.546.3S95dr
http://www.redis.cn/topics/introduction.html
http://www.redis.cn/topics/partitioning.html
二：分佈式緩存tair
https://blog.csdn.net/farphone/article/details/53522383
三：分佈式消息kafka（文件形式存儲硬盤）
https://help.aliyun.com/document_detail/68156.html?spm=a2c4g.11186623.6.543.IG04U8
https://blog.csdn.net/qq_21989939/article/details/79452065
生產建議2個集羣【1個業務1個日誌集羣，每個集羣3臺機器（建議配置4c8g2t）】，測試可一臺、kafka默認不支持ons的tag標籤，但通過透傳已經支持，
巨量的partition不支持
集羣上部署kafka代碼一模一樣，但數據是分佈式存儲，即每臺機器上的partition存儲數據不一樣
kafka比ons性能高大概3倍，因爲從代碼上進行了優化，包括但不限於線程和nio模式、跑批處理
kafka的延遲和事務的問題
topic與partition 是一對多，一個partition 在磁盤是一個文件
Data written to Kafka is written to disk and replicated for fault-tolerance. Kafka allows producers to wait on acknowledgement so that a write isn't considered complete until it is fully replicated and guaranteed to persist even if the server written to fails.
Kafka uses ZooKeeper so you need to first start a ZooKeeper server if you don't already have one.
zookeeper註冊節點和topic信息 【http://kafka.apache.org/documentation/】
Kafka的Message存儲採用了分區(partition)，分段(LogSegment)和稀疏索引這幾個手段來達到了高效性