[
https://issues.apache.org/jira/browse/IGNITE-24753?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18108914#comment-18108914
]
Ignite TC Bot commented on IGNITE-24753:
----------------------------------------
I reviewed implementation PR #11955 and commit 1bf7fdb123 in the local Ignite 2
checkout. This fixes a Calcite MergeJoin execution freeze affecting TPC-H query
execution and does not introduce new SQL syntax, public API, configuration, or
user-facing documentation surface. I removed only Docs Required and preserved
Release Notes Required.
> Calcite. TPC-H query #16: too long for scale=0.1 and 1.0
> --------------------------------------------------------
>
> Key: IGNITE-24753
> URL: https://issues.apache.org/jira/browse/IGNITE-24753
> Project: Ignite
> Issue Type: Bug
> Reporter: Sergey Korotkov
> Assignee: Aleksey Plekhanov
> Priority: Major
> Labels: ise, tpch
> Fix For: 2.18
>
> Attachments: TpchQ16Test.java
>
> Time Spent: 1h 10m
> Remaining Estimate: 0h
>
> Works for scale factor 0.01 but doesn't complete in 5 minutes for scale
> factors 0.1 and 1.0.
> ***
> Load test (1 client node, 3 server nodes) shows that there is an
> unproportional increase of HEAP memory usage for scales 0.1 and 1.0 comparing
> to 0.01.
> For scale=0.01 it were executed about 700 queries in 5 seconds. With average
> 2.99Mb and 5.68Mb of heap allocation per request (on client and server
> correspondingly).
> For scales 0.1 and 1.0 single query was calcelled after 5 minutes, having
> Gigabytes of allocated per the single query.
>
> | ** | ** |*{color:#000000}Scale = 0.01{color}*|*{color:#000000}Scale =
> 0.1{color}*|*{color:#000000}Scale = 1.0{color}*|
> |*{color:#000000}benchmark{color}*|*{color:#000000}metric{color}*|*{color:#000000}value{color}*|*{color:#000000}value{color}*|*{color:#000000}value{color}*|
> |{color:#000000}Q16{color}|{color:#000000}client_heap_inTestAllocatedPerCall_MB/call_mean{color}|{color:#000000}2.99{color}|{color:#000000}11,390.24{color}|{color:#000000}11,495.04{color}|
> |{color:#000000}Q16{color}|{color:#000000}client_heap_inTestTotalAllocated_MB_mean{color}|{color:#000000}2,071.28{color}|{color:#000000}11,390.24{color}|{color:#000000}11,495.04{color}|
> |{color:#000000}Q16{color}|{color:#000000}server_heap_inTestAllocatedPerCall_MB/call_mean{color}|{color:#000000}5.68{color}|{color:#000000}15,489.86{color}|{color:#000000}8,576.07{color}|
> |{color:#000000}Q16{color}|{color:#000000}server_heap_inTestTotalAllocated_MB_mean{color}|{color:#000000}3,939.98{color}|{color:#000000}15,489.86{color}|{color:#000000}8,576.07{color}|
> |{color:#000000}Q16{color}|{color:#000000}count{color}|{color:#000000}693.33{color}|{color:#000000}1.00{color}|{color:#000000}1.00{color}|
> |{color:#000000}Q16{color}|{color:#000000}failed_count{color}|{color:#000000}0.00{color}|{color:#000000}1.00{color}|{color:#000000}1.00{color}|
> ***
> Performance of the TPC-H queries can also be tested locally in JMH test at
> [https://github.com/apache/ignite/blob/master/modules/benchmarks/src/main/java/org/apache/ignite/internal/benchmarks/jmh/sql/tpch/TpchBenchmark.java]
>
> ***
> See something similar for the query #19 at IGNITE-24756
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
