PIOUS (Parallel Input/OUtput System) is a parallel file system I developed long ago for the PVM (Parallel Virtual Machine) cluster computing environment. PVM was widely used for (what we now call) Big Data problems before the days of MPI and Hadoop.
The PIOUS architecture is very similar to that of the Hadoop file system (HDFS). Applications link with the PIOUS library and access files declustered across a set of PIOUS data servers (PDS). The PIOUS service coordinator (PSC) supports coordination and metadata operations but is not on the data path.
While the PIOUS architecture is similar to HDFS the feature set is of course different. PIOUS focuses on multi-dimensional files and light-weight concurrency control to support parallel file access.
PIOUS is also written in C whereas HDFS is in Java.
So why am I putting this code out on GitHub when PVM is no longer widely used? Good question.
Mainly because PIOUS has minimal dependence on PVM and can be trivially ported to other environments. All PVM dependencies are isolated to a single file: src/pdce/pdce.c (and related .h files in that directory).
So if someone is looking for HDFS-like functionality but in C rather than Java, it will be a lot easier to port PIOUS and use it as a starting point versus starting from scratch.
I confirmed that both PIOUS and PVM still build, run, and pass tests on recent Linux systems. Complete instructions for doing so can be found in doc/piousUG1.2.pdf.