Judy looks interesting for all memory applications.
How would Judy perform on-disk? Would it be trivial to
create an on-disk Judy format? Perhaps an SQL RDBMS
could use such an extension.
An on-disk format would have greater design
implications than an in-memory implementation.
Remember that sparseness on disk is bad, as gaps of
unused data eat disk space. Fragmentation algorithms
would need to be examined. Entries could easily exist
as arrays of arrays of locations which got concatonated
together to form the entry; but massive fragmentation
would still be bad.
Perhaps Judy can work through its 1.x versions, and
slate an on-disk format for consideration after 1.x is
proper in memory, as a goal for 2.0.
Log in to post a comment.