What is the main goal of Multiprocess ?

Definitely go for option 2.

Option 1 is the classic disk PLC mechanism, which is less reliable and only redundant up to 2 I/O servers.
For the redundancy to work you need to share a folder on each I/O server and point the CDK files to both folders, which is unreliable.
Also, a CDK file has a proprietary binary format which does not allow you to see the values of the tags in it.

Option 2 is the "new" Persisant Memory mechanism, which uses standard Citect server-server communication channels and supports redundancy for all I/O servers in your system.
The XML files are plain text files which can be useful to debug / inspect tag values.

The reason why OPC (or OFSOPC) is suggested for the protocol is that you don't need to specify Variable Tag addresses like you would with the GENERIC protocol (D1, I1, S1, etc.).

Definitely go for option 2.

Option 1 is the classic disk PLC mechanism, which is less reliable and only redundant up to 2 I/O servers.
For the redundancy to work you need to share a folder on each I/O server and point the CDK files to both folders, which is unreliable.
Also, a CDK file has a proprietary binary format which does not allow you to see the values of the tags in it.

Option 2 is the "new" Persisant Memory mechanism, which uses standard Citect server-server communication channels and supports redundancy for all I/O servers in your system.
The XML files are plain text files which can be useful to debug / inspect tag values.

The reason why OPC (or OFSOPC) is suggested for the protocol is that you don't need to specify Variable Tag addresses like you would with the GENERIC protocol (D1, I1, S1, etc.).