An Eventually Consistent Data Model For Erlang (and Riak)

CAP理论指出：一个分布式系统不可能同时满足一致性(Consistency)、可用性(Availibility)和分区容忍性(Partition Tolerance)这三个需求,最多只能同时满足其中的两个。在Riak强调的是Availibility和Partition Tolerance,它的Consistency是指eventually consistent，在一些concurrency的场合，你会发现你的data model会有inconsistent的情况出现，可以使用statebox去解决这个问题，下面用statebox_riak去实现博客互粉的功能：

statebox_riak的github地址:点击打开链接

riakc_pb_socket:set_bucket(Pid, ?BUCKET, [{allow_mult, true}])

{ok, Pid} = riakc_pb_socket:start_link("127.0.0.1", 8087),
S = statebox_riak:new([{riakc_pb_socket, Pid}, {max_queue, 20}]),
statebox_riak:apply_bucket_ops(?BUCKET,
    [
        {[list_to_binary(FromId)], statebox_orddict:f_union(following, [list_to_binary(ToId)])},
        {[list_to_binary(ToId)], statebox_orddict:f_union(followers, [list_to_binary(FromId)])}
    ], S),
riakc_pb_socket:stop(Pid).

更多信息可以看看这篇文章 点击打开链接