How is the "friend Dynamics" stream of 100,000-level active users efficiently implemented?

For a 100,000-level active user social network, how does MySQL and Redis database design for the user's homepage stream (which is the friend dynamic)? Are there any requirements for PHP?
Note: limited to single machine, or two servers. Large clusters are not considered.

In addition, I would like to know how Sina Weibo this tens active user's friend Dynamic Data sheet is implemented?

Reply content:

For a 100,000-level active user social network, how does MySQL and Redis database design for the user's homepage stream (which is the friend dynamic)? Are there any requirements for PHP?
Note: limited to single machine, or two servers. Large clusters are not considered.

In addition, I would like to know how Sina Weibo this tens active user's friend Dynamic Data sheet is implemented?

Similar to the real-time information flow, such as microblogging, need to have a sound cache system, everyone's information flow is directly pushed over, of course, according to some user policies can also be pulled by the way. This will ensure timeliness.
100,000 level 2 can be enough.

Redis+mysql, hot and cold data stored separately, high-frequency reading data memory, should be anti-millions of users no problem. Now there is a high-end fusion IO solution, SSD read and write performance has been greatly improved. Weibo should adopt a poll rather than push strategy.

