[Apue] some questions multiple processes to read and write pipeline

For one-pipe:

1) After the write process close the write pipeline, the process continues to read read read pipeline will lead to the return 0;

2) After the reading process closed pipeline read, write, write pipeline process continues to inspire SIGPIPE signal, if captured, the write return -1;

For many-to-pipe:

1) multiple write-write pipeline process shuts down, the process continues to read read read pipeline will lead to the return 0; (ibid)

2) After more reading process piping close reading, writing process and the pipeline continues to write SIGPIPE signal does not excite, but to write a successful return. (Test no large amounts of data, so a small amount of data)

doubt:

What is the multiple processes and a shared pipeline process in the kernel similarities and differences, why the performance will be different?