Field-stripe encoding similar in spirit to that of Google's Dremel. As is common in many publications, the "what" and "why" were omitted for lack of time and space from the Dremel paper. We wanted to understand the requirements, assumptions, trade-offs, etc. necessary for creating a field-stripe based encoding and compare that with the results / implementation presented in the Dremel paper.
The primary artifact is the document Efficient Field-Striped, Nested, Disk-backed Record Storage. The document is by no means complete and is in desperate need of a re-write but it is packed with information. In order to give the document some teeth and to understand the implications of what it proposes, a simple Java-based implementation was crafted.
The software has been updated to use Protostuff as a parser.
It is our hope that this work, although incomplete, can be used to "prime the pump" for an open source implementation of a field-stripe backed system (such as Drill).