GitHub user jianqiao opened a pull request:
https://github.com/apache/incubator-quickstep/pull/172
Query optimization with ExactFilter
This is a follow-up optimization based on the facility provided by
LIPFilters. Note that LIP (lookahead information passing) is an optimization
that we can inject efficient filters (e.g. bloom filters) into
Select/HashJoin/Aggregate operators to pre-filter the input relations.
This PR strength-reduces `HashJoin`s (including inner/semi/anti joins) into
`FilterJoin`s. The semantics of a `FilterJoin` is simple: if certain conditions
are met, we can build a bit vector from the build side and use the bit vector
to _filter_ the probe side.
The execution part is slightly more optimized: a `FilterJoin` will not
always be converted into a `SelectOperator` plus a `LIPFilter` as its semantics
indicates. Instead, in most situations we can avoid creating the
`SelectOperator` by attaching the `LIPFilter` properly to some downstream
operators â thus avoid unnecessary materialization of intermediate relations.
Below shows the performance improvement for SSB scale factor 100 on a
cloudlab machine:
**SSB SF100**|**master (ms)**|**w/ ExactFilter (ms)**
:-----:|:-----:|:-----:
Q01|709|574
Q02|648|593
Q03|605|564
Q04|906|675
Q05|754|457
Q06|498|549
Q07|1687|1696
Q08|598|591
Q09|481|470
Q10|450|442
Q11|1208|882
Q12|876|656
Q13|515|475
Total|9937|8625
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/apache/incubator-quickstep exact-filter
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/incubator-quickstep/pull/172.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #172
----
----
---
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 infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
