Github user DaveBirdsall commented on a diff in the pull request:
https://github.com/apache/incubator-trafodion/pull/1206#discussion_r133313356
--- Diff: core/sql/sqlcomp/nadefaults.cpp ---
@@ -1759,6 +1763,7 @@ SDDkwd__(EXE_DIAGNOSTIC_EVENTS, "OFF"),
// exposure.
DDkwd__(HBASE_DELETE_COSTING, "ON"),
DDflt0_(HBASE_DOP_PARALLEL_SCANNER, "0."),
+ DDkwd__(HBASE_ESTIMATE_ROW_COUNT_VIA_COPROCESSOR, "OFF"),
--- End diff --
No. In my tests, there was a small performance penalty for using the
coprocessor. We must have the coprocessor if HBase encryption is being used,
but otherwise we can just use the client code.
Your next question might be: why is there a performance penalty? I looked
into that a little bit. I think it is because we do dynamic coprocessor
loading. It may be that the penalty is not real: this coprocessor is packaged
with the DTM coprocessors. I think they all get loaded together. Assuming this
is true, we are simply paying this penalty earlier (at estimate row count time)
rather than at transaction start time.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
