Data Access from the LAT
Assumptions
- apply queries to ntuples and produce output as FITS, trees or ntuples
- Root I/O has features we should be able to exploit
- read only branches involved in the query before selecting successful
rows
- should be "free" to carry many ntuple columns
- only paying disk space, not access time
- random access to events (rows) inside files
- can index files internally to attach event ID to row number in the files
- extract tree events from ntuple index
- probably need to put ntuples on the query server for best access
- if needed, we can investigate Root's PROOF parallel query system
- automagically breaks up and routes queries to different nodes and
reassembles the answer before returning them to the user
Links
R.Dubois Last Modified:
2003-10-20 20:53:39 -0700