Details
-
Bug
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
Impala 1.3.1
-
None
-
None
Description
This is the catalog object that scales most quickly (one per replica) and we've made it pretty heavyweight by storing the TNetworkAddress of the host it is on.
This should instead use the same technique that we used to reduce the catalog size (i.e. index into list of hosts) or, better, removing this all together since it contains a subset and more expensive version of what is stored in THdfsFileBlock.
This will likely speedup our query start up noticeably.