From my recent findings, it looks like the MPIO project in the dm layer has come a long way towards making Linux an enterprise-class OS. Lately, I’ve been curious in regards to the roots of the MPIO project as an alternative to the costly and sometimes buggy EMC Powerpath. Can anyone shed some light on the following questions?
When this project started, what were the goals? Since there are so many places where an IO can fail, were specific IO errors targeted or was development limited to IO errors caused only by physical path loss? Was there a specific reliability target (i.e. – twice as reliable as a single path to IO target)? Also--this is a more general question--where are all the places an IO can fail in Linux?
I look forward to learning more about this project. Thanks in advance for any replies!