davepacheco
commented
10 years ago This is implemented, but still pretty naive. It doesn't ever clean up files from the cache; it doesn't parallelize fetching and querying; and it always checks Manta for the list of indexes, rather than servicing the request entirely locally if we can tell that there are no missing files.
To avoid the overhead of running compute jobs for queries, it would be nice to have a mode of querying a local cache of indexes whose canonical copies are stored in Manta. It would look just like "query-mjob" except that it would fetch the index files to a local cache and query them like "query-tree" does. Subsequent runs should be much faster.