You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues-all@impala.apache.org by "ASF subversion and git services (Jira)" <ji...@apache.org> on 2023/06/08 05:35:00 UTC
[jira] [Commented] (IMPALA-4530) Sort node after exchange should start sorting after first RowBatch is received
[ https://issues.apache.org/jira/browse/IMPALA-4530?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17730387#comment-17730387 ]
ASF subversion and git services commented on IMPALA-4530:
---------------------------------------------------------
Commit 80c1d2dbaabc78bdf1b6e4da5475bfa365cd375e in impala's branch refs/heads/master from noemi
[ https://gitbox.apache.org/repos/asf?p=impala.git;h=80c1d2dba ]
IMPALA-4530: Implement in-memory merge of quicksorted runs
This change aims to decrease back-pressure in the sorter. It offers an
alternative for the in-memory run formation strategy and sorting
algorithm by introducing a new in-memory merge level between the
in-memory quicksort and the external merge phase.
Instead of forming one big run, it produces many smaller in-memory runs
(called miniruns), sorts those with quicksort, then merges them
in memory, before spilling or serving GetNext().
The external merge phase remains the same.
Works with MAX_SORT_RUN_SIZE development query option that determines
the maximum number of pages in a 'minirun'. The default value of
MAX_SORT_RUN_SIZE is 0, which keeps the original implementation of 1
big initial in-memory run. Other options are integers of 2 and above.
The recommended value is 10 or more, to avoid high fragmentation
in case of large workloads and variable length data.
Testing:
- added MAX_SORT_RUN_SIZE as an additional test dimension to
test_sort.py with values [0, 2, 20]
- additional partial sort test case (inserting into partitioned
kudu table)
- manual E2E testing
Change-Id: I58c0ae112e279b93426752895ded7b1a3791865c
Reviewed-on: http://gerrit.cloudera.org:8080/18393
Reviewed-by: Impala Public Jenkins <im...@cloudera.com>
Reviewed-by: Csaba Ringhofer <cs...@cloudera.com>
Tested-by: Csaba Ringhofer <cs...@cloudera.com>
> Sort node after exchange should start sorting after first RowBatch is received
> ------------------------------------------------------------------------------
>
> Key: IMPALA-4530
> URL: https://issues.apache.org/jira/browse/IMPALA-4530
> Project: IMPALA
> Issue Type: Bug
> Components: Backend
> Reporter: Mostafa Mokhtar
> Assignee: Noemi Pap-Takacs
> Priority: Major
> Labels: performance
>
> Sort node after exchange doesn't start sorting until all data is received which add lots of latency to the query.
> Not clear if this optimization would still make sense for a Scan followed by a sort run using the same thread.
> Query
> {code}
> insert into tpcds_1000_parquet.store_sales_insert partition(ss_sold_date_sk, ss_quantity) /*+ clustered*/
> select
> ss_sold_time_sk,
> ss_item_sk ,
> ss_customer_sk,
> ss_cdemo_sk,
> ss_hdemo_sk,
> ss_addr_sk,
> ss_store_sk,
> ss_promo_sk,
> ss_ticket_number ,
> ss_wholesale_cost ,
> ss_list_price ,
> ss_sales_price ,
> ss_ext_discount_amt ,
> ss_ext_sales_price ,
> ss_ext_wholesale_cost ,
> ss_ext_list_price ,
> ss_ext_tax ,
> ss_coupon_amt ,
> ss_net_paid ,
> ss_net_paid_inc_tax ,
> ss_net_profit,
> ss_sold_date_sk , ss_quantity
> from store_sales
> {code}
> Plan
> {code}
> WRITE TO HDFS [tpcds_1000_parquet.store_sales_insert, OVERWRITE=false, PARTITION-KEYS=(ss_sold_date_sk,ss_quantity)]
> | partitions=180576
> | hosts=15 per-host-mem=17.88GB
> |
> 02:SORT
> | order by: ss_sold_date_sk DESC NULLS LAST, ss_quantity DESC NULLS LAST
> | hosts=15 per-host-mem=1.45GB
> | tuple-ids=1 row-size=100B cardinality=2879987999
> |
> 01:EXCHANGE [HASH(ss_sold_date_sk,ss_quantity)]
> | hosts=15 per-host-mem=0B
> | tuple-ids=0 row-size=100B cardinality=2879987999
> |
> 00:SCAN HDFS [tpcds_1000_parquet.store_sales, RANDOM]
> partitions=1824/1824 files=1824 size=189.24GB
> table stats: 2879987999 rows total
> column stats: all
> hosts=15 per-host-mem=88.00MB
> tuple-ids=0 row-size=100B cardinality=2879987999
> {code}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-all-unsubscribe@impala.apache.org
For additional commands, e-mail: issues-all-help@impala.apache.org