key实现的方法
- encoding
- 提前把key都生成
Data partition:
- Range based
- Hash based
Caching, and pruning
------------------
Designing Pastebin, 非常类似tinyUrl
-------------------------
Designing Instagram
Photo storage, object storage
News feed, pulling / pushing model
--------------------
Designing Dropbox
Divide file into chunks, transmit less data during synchronization
local client: long polling to the server to check changes
Updates are frequent, use queue between server and client
only transmitting the diff between files
only transmitting the diff between files
先传data，再更新metadata
2种dedup的方法
metadata partition的方法
--------------
Designing Facebook Messenger
request/response workflow要看看
每个client都long pulling的话，一共得开多少个socket
保证message的顺序：server端的timestamp ＋ client本地加sequence
用wild column data， hbase。 为什么要用这个？
用户的online／offline status的处理，client pull partial users
Push notification when a user is off
------------------
Designing Twitter
Data sharding - 用timestamp和tweetid的组合, 如何解决？
Cache last 3 days of tweets for one user
---------------
Designing Youtube or Netflix
thumbnail 的读写
processing queue for uploading and encoding videos
2种dedup的方法
metadata partition的方法
--------------
Designing Facebook Messenger
request/response workflow要看看
每个client都long pulling的话，一共得开多少个socket
保证message的顺序：server端的timestamp ＋ client本地加sequence
用wild column data， hbase。 为什么要用这个？
用户的online／offline status的处理，client pull partial users
Push notification when a user is off
------------------
Designing Twitter
Data sharding - 用timestamp和tweetid的组合, 如何解决？
Cache last 3 days of tweets for one user
---------------
Designing Youtube or Netflix
thumbnail 的读写
processing queue for uploading and encoding videos
