Aman Sinha has posted comments on this change. ( http://gerrit.cloudera.org:8080/15683 )
Change subject: IMPALA-3741 [part 2]: Push runtime bloom filter to Kudu ...................................................................... Patch Set 3: (1 comment) http://gerrit.cloudera.org:8080/#/c/15683/3/common/thrift/PlanNodes.thrift File common/thrift/PlanNodes.thrift: http://gerrit.cloudera.org:8080/#/c/15683/3/common/thrift/PlanNodes.thrift@124 PS3, Line 124: BLOOM_MIN_MAX = 2 In the case of both Bloom and Min-Max filters, is the order of evaluation determined by Kudu ? For columns that are not sorted (partially sorted within each tablet), the Min-Max filters could potentially not eliminate much and be wasted effort, so Bloom filter should be evaluated first. On the other hand, for sorted columns, it would make sense to apply Min-Max filters first since Bloom filter has false positives. It seems that for a specific column, creating two types of filters would incur overhead during execution - since both have to be sent to coordinator and broadcast to executors. Any thoughts on that ? -- To view, visit http://gerrit.cloudera.org:8080/15683 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: Impala-ASF Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: I9100076f68ea299ddb6ec8bc027cac7a47f5d754 Gerrit-Change-Number: 15683 Gerrit-PatchSet: 3 Gerrit-Owner: Wenzhe Zhou <[email protected]> Gerrit-Reviewer: Aman Sinha <[email protected]> Gerrit-Reviewer: Bankim Bhavsar <[email protected]> Gerrit-Reviewer: Impala Public Jenkins <[email protected]> Gerrit-Reviewer: Thomas Tauber-Marshall <[email protected]> Gerrit-Reviewer: Wenzhe Zhou <[email protected]> Gerrit-Comment-Date: Fri, 10 Apr 2020 02:08:39 +0000 Gerrit-HasComments: Yes
