Repository: impala Updated Branches: refs/heads/master 1e79f1479 -> 87be63e32
IMPALA-6821: Push down limits into Kudu This patch takes advantage of a recent change in Kudu (KUDU-16) that exposes the ability to set limits on KuduScanners. Since each KuduScanner corresponds to a scan token, and there will be multiple scan tokens per query, this is just a performance optimization in cases where the limit is smaller than the number of rows per token, and Impala still needs to apply the limit on our side for cases where the limit is greater than the number of rows per token. Testing: - Added e2e tests for various situations where limits are applied at a Kudu scan node. - For the query 'select * from tpch_kudu.lineitem limit 1', a best case perf scenario for this change where the limit is highly effective, the time spent in the Kudu scan node was reduced from 6.107ms to 3.498ms (avg over 3 runs). - For the query 'select count(*) from (select * from tpch_kudu.lineitem limit 1000000) v', a worst case perf scenario for this change where the limit is ineffective, the time spent in the Kudu scan node was essentially unchanged, 32.815ms previously vs. 29.532ms (avg over 3 runs). Change-Id: Ibe35e70065d8706b575e24fe20902cd405b49941 Reviewed-on: http://gerrit.cloudera.org:8080/10119 Reviewed-by: Thomas Tauber-Marshall <tmarsh...@cloudera.com> Tested-by: Impala Public Jenkins <impala-public-jenk...@cloudera.com> Project: http://git-wip-us.apache.org/repos/asf/impala/repo Commit: http://git-wip-us.apache.org/repos/asf/impala/commit/87be63e3 Tree: http://git-wip-us.apache.org/repos/asf/impala/tree/87be63e3 Diff: http://git-wip-us.apache.org/repos/asf/impala/diff/87be63e3 Branch: refs/heads/master Commit: 87be63e321f688486b98d4ea69200967a8a2effa Parents: 1e79f14 Author: Thomas Tauber-Marshall <tmarsh...@cloudera.com> Authored: Fri Apr 13 20:38:16 2018 +0000 Committer: Impala Public Jenkins <impala-public-jenk...@cloudera.com> Committed: Fri Apr 27 21:55:11 2018 +0000 ---------------------------------------------------------------------- be/src/exec/kudu-scanner.cc | 5 + .../queries/QueryTest/kudu_limit.test | 112 +++++++++++++++++++ tests/query_test/test_kudu.py | 3 + 3 files changed, 120 insertions(+) ---------------------------------------------------------------------- http://git-wip-us.apache.org/repos/asf/impala/blob/87be63e3/be/src/exec/kudu-scanner.cc ---------------------------------------------------------------------- diff --git a/be/src/exec/kudu-scanner.cc b/be/src/exec/kudu-scanner.cc index 3bc4441..9e90bdb 100644 --- a/be/src/exec/kudu-scanner.cc +++ b/be/src/exec/kudu-scanner.cc @@ -226,6 +226,11 @@ Status KuduScanner::OpenNextScanToken(const string& scan_token, bool* eos) { } } + if (scan_node_->limit() != -1 && conjunct_evals_.empty()) { + KUDU_RETURN_IF_ERROR(scanner_->SetLimit(scan_node_->limit()), + "Failed to set limit on scan."); + } + { SCOPED_TIMER(state_->total_storage_wait_timer()); KUDU_RETURN_IF_ERROR(scanner_->Open(), "Unable to open scanner"); http://git-wip-us.apache.org/repos/asf/impala/blob/87be63e3/testdata/workloads/functional-query/queries/QueryTest/kudu_limit.test ---------------------------------------------------------------------- diff --git a/testdata/workloads/functional-query/queries/QueryTest/kudu_limit.test b/testdata/workloads/functional-query/queries/QueryTest/kudu_limit.test new file mode 100644 index 0000000..c72afc9 --- /dev/null +++ b/testdata/workloads/functional-query/queries/QueryTest/kudu_limit.test @@ -0,0 +1,112 @@ +==== +---- QUERY +# limit 0 +select * from functional_kudu.alltypes limit 0 +---- RESULTS +---- TYPES +int,boolean,tinyint,smallint,int,bigint,float,double,string,string,timestamp,int,int +==== +---- QUERY +# no predicate, nondeterministic (no order by) so only check number of rows returned +select count(*) from (select * from functional_kudu.alltypes limit 2) v +---- RESULTS +2 +---- TYPES +bigint +==== +---- QUERY +# no predicate, deterministic (limit doesn't exclude any rows) so check actual values +select * from functional_kudu.alltypestiny limit 100 +---- RESULTS : VERIFY_IS_EQUAL_SORTED +0,true,0,0,0,0,0,0,'01/01/09','0',2009-01-01 00:00:00,2009,1 +2,true,0,0,0,0,0,0,'02/01/09','0',2009-02-01 00:00:00,2009,2 +4,true,0,0,0,0,0,0,'03/01/09','0',2009-03-01 00:00:00,2009,3 +1,false,1,1,1,10,1.100000023841858,10.1,'01/01/09','1',2009-01-01 00:01:00,2009,1 +5,false,1,1,1,10,1.100000023841858,10.1,'03/01/09','1',2009-03-01 00:01:00,2009,3 +6,true,0,0,0,0,0,0,'04/01/09','0',2009-04-01 00:00:00,2009,4 +7,false,1,1,1,10,1.100000023841858,10.1,'04/01/09','1',2009-04-01 00:01:00,2009,4 +3,false,1,1,1,10,1.100000023841858,10.1,'02/01/09','1',2009-02-01 00:01:00,2009,2 +---- TYPES +int,boolean,tinyint,smallint,int,bigint,float,double,string,string,timestamp,int,int +==== +---- QUERY +# Kudu predicate on PK col, nondeterministic +select count(id) from (select * from functional_kudu.alltypes where id > 0 limit 3) v +---- RESULTS +3 +---- TYPES +bigint +==== +---- QUERY +# Kudu predicate on PK col, deterministic +select * from functional_kudu.alltypestiny where id > 4 limit 3 +---- RESULTS : VERIFY_IS_EQUAL_SORTED +5,false,1,1,1,10,1.100000023841858,10.1,'03/01/09','1',2009-03-01 00:01:00,2009,3 +6,true,0,0,0,0,0,0,'04/01/09','0',2009-04-01 00:00:00,2009,4 +7,false,1,1,1,10,1.100000023841858,10.1,'04/01/09','1',2009-04-01 00:01:00,2009,4 +---- TYPES +int,boolean,tinyint,smallint,int,bigint,float,double,string,string,timestamp,int,int +==== +---- QUERY +# Kudu predicate on non-PK col, nondeterministic +select count(tinyint_col) from (select * from functional_kudu.alltypes where tinyint_col = 6 limit 4) v +---- RESULTS +4 +---- TYPES +bigint +==== +---- QUERY +# Kudu predicate on non-PK col, deterministic +select * from functional_kudu.alltypestiny where tinyint_col = 1 limit 4 +---- RESULTS : VERIFY_IS_EQUAL_SORTED +1,false,1,1,1,10,1.100000023841858,10.1,'01/01/09','1',2009-01-01 00:01:00,2009,1 +5,false,1,1,1,10,1.100000023841858,10.1,'03/01/09','1',2009-03-01 00:01:00,2009,3 +7,false,1,1,1,10,1.100000023841858,10.1,'04/01/09','1',2009-04-01 00:01:00,2009,4 +3,false,1,1,1,10,1.100000023841858,10.1,'02/01/09','1',2009-02-01 00:01:00,2009,2 +---- TYPES +int,boolean,tinyint,smallint,int,bigint,float,double,string,string,timestamp,int,int +==== +---- QUERY +# Impala predicate, nondeterministic +select count(string_col) from (select * from functional_kudu.alltypes where id % 2 = 0 limit 2) v +---- RESULTS +2 +---- TYPES +bigint +==== +---- QUERY +# Impala predicate, deterministic +select * from functional_kudu.alltypestiny where id % 2 = 0 limit 100 +---- RESULTS : VERIFY_IS_EQUAL_SORTED +0,true,0,0,0,0,0,0,'01/01/09','0',2009-01-01 00:00:00,2009,1 +2,true,0,0,0,0,0,0,'02/01/09','0',2009-02-01 00:00:00,2009,2 +4,true,0,0,0,0,0,0,'03/01/09','0',2009-03-01 00:00:00,2009,3 +6,true,0,0,0,0,0,0,'04/01/09','0',2009-04-01 00:00:00,2009,4 +---- TYPES +int,boolean,tinyint,smallint,int,bigint,float,double,string,string,timestamp,int,int +==== +---- QUERY +# Both Impala and Kudu predicates, nondeterministic +select count(month) from + (select * from functional_kudu.alltypes where id % 2 = 0 and id > 4 limit 5) v +---- RESULTS +5 +---- TYPES +bigint +==== +---- QUERY +# Both Impala and Kudu predicates, deterministic +select * from functional_kudu.alltypestiny where id % 2 = 0 and id > 4 limit 5 +---- RESULTS +6,true,0,0,0,0,0,0,'04/01/09','0',2009-04-01 00:00:00,2009,4 +---- TYPES +int,boolean,tinyint,smallint,int,bigint,float,double,string,string,timestamp,int,int +==== +---- QUERY +# large limit, nondeterministic +select count(*) from (select * from functional_kudu.alltypes where id % 2 = 0 and id > 1 limit 1000) v; +---- RESULTS +1000 +---- TYPES +bigint +==== http://git-wip-us.apache.org/repos/asf/impala/blob/87be63e3/tests/query_test/test_kudu.py ---------------------------------------------------------------------- diff --git a/tests/query_test/test_kudu.py b/tests/query_test/test_kudu.py index b0ced46..ae76f75 100644 --- a/tests/query_test/test_kudu.py +++ b/tests/query_test/test_kudu.py @@ -107,6 +107,9 @@ class TestKuduOperations(KuduTestSuite): def test_kudu_describe(self, vector, unique_database): self.run_test_case('QueryTest/kudu_describe', vector, use_db=unique_database) + def test_kudu_limit(self, vector, unique_database): + self.run_test_case('QueryTest/kudu_limit', vector, use_db=unique_database) + def test_kudu_column_options(self, cursor, kudu_client, unique_database): """Test Kudu column options""" encodings = ["ENCODING PLAIN_ENCODING", ""]