As the original author and core developer of John the Ripper, a password security auditing tool (password cracker), I am happy that Rapid7 selected our project for Magnificent7. This is not the first time for Rapid7 to support John the Ripper development. As some of you recall, we've optimized the DES S-box expressions and made other DES-related changes with Rapid7's support last year.
This time, under the Magnificent7 program, we're going to enhance John the Ripper's parallel and distributed processing capabilities.
Currently, John the Ripper is able to use multiple CPUs/cores on one machine by means of OpenMP, but this is limited to a subset of hash/cipher types. John the Ripper is also able to use multiple nodes on a network, but this requires an MPI setup. The plan is to eliminate these limitations.
For parallel processing on one machine, the plan is to add built-in forking of subprocesses. This will be compatible with all formats. (The existing OpenMP support will continue to evolve as well.)
For distributed processing, it is planned to add both a simple command-line option usable to split the workload between multiple disconnected nodes, as well as a full-blown implementation with networking. (There will be no dependency on MPI in either case. The existing MPI support will stay as well.)
Also planned is a revision to the incremental mode. It is expected to provide greater efficiency (more passwords cracked per N candidates tried). At the same time we'll start providing .chr files for lengths beyond 8 (this is already supported, but is currently not the default).
These enhancements have consistently stayed on the back burner, but the funding from Magnificent7 will change that. Thank you, Rapid7, for running this magnificent program.