about cluster strategy

Jun Liu
Added almost 4 years ago


I just go through the default cluster strategy, from the paper, I know that each server is "equivalent", the main problem is the server would hold the data of one user, so it need to find a way to route the data to the right server, I am thinking about whether we can store these data in distributed storage, so that each server can share the data, just like most website use memcache to share the session, then each server would be truly equivalent, and would be more scalable

I need to implement our own cluster strategy, I am thinking about two possible ways:

  1. just as I say, use distributed storage to share the data

  2. similar to the default strategy, the recerver would route the package to the right server, the difference is I will store the id and server map somewhere, so that it doesn't need to send one by one

can some one provide any suggestions?

Thanks in advance