io buffer sizes

So doing some research I found a bunch of old 2007-2010 article talking about io buffers (read_buffer_size and others) and the performance change around 128k when they change from molloc to mmap. I was wondering if there is any updated info on this? And if in virtualized environments this changes at all etc. I would love to see a new blog that deep dives into this.

http://inaugust.com/post/15
http://www.percona.com/blog/2007/09/17/mysql-what-read_buffer_size-value-is-optimal/
http://www.percona.com/blog/2010/10/25/impact-of-the-sort-buffer-size-in-mysql/