[
https://issues.apache.org/jira/browse/PHOENIX-6647?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17490481#comment-17490481
]
Lars Hofhansl commented on PHOENIX-6647:
----------------------------------------
Now that I spent a 1/2 looking at this, the cost estimation is all weird anyway.
Why do we apply a parallelLevel for join and sort operations? We do not apply
that to scans, which also run in parallel on the servers. If anything we might
want to apply a multiplier for client sort/joins instead.
That throws the calculations off. In addition we're already mixing CPU, IO, and
memory metrics in the code, but then say we are not using CPU and memory, yet.
I doubt anyone has the cost based optimizer enabled since it is default off. So
I wonder if there's any value spending a few more hours on this...
> A local index should not be chosen for a full scan if that scan is not
> covered by the index.
> --------------------------------------------------------------------------------------------
>
> Key: PHOENIX-6647
> URL: https://issues.apache.org/jira/browse/PHOENIX-6647
> Project: Phoenix
> Issue Type: Bug
> Affects Versions: 5.1.2
> Reporter: Lars Hofhansl
> Priority: Major
> Attachments: 6647-5.1.txt, 6647-v2-5.1.txt
>
>
> {code}
> > explain select * from lineitem;
> +-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+----------------+-----+
> |
> PLAN
> | EST_BYTES_READ | EST |
> +-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+----------------+-----+
> | CLIENT 103-CHUNK 17711182 ROWS 1059064693 BYTES PARALLEL 2-WAY ROUND ROBIN
> RANGE SCAN OVER LINEITEM [1]
> | 1059064693 | 177 |
> | SERVER MERGE [0.PARTKEY, 0.SUPPKEY, 0.QUANTITY, 0.EXTENDEDPRICE,
> 0.DISCOUNT, 0.TAX, 0.RETURNFLAG, 0.LINESTATUS, 0.COMMITDATE, 0.RECEIPTDATE,
> 0.SHIPINSTRUCT, 0.SHIPMODE, 0.COMMENT] | 1059064693 | 177 |
> | SERVER FILTER BY FIRST KEY ONLY
>
> | 1059064693 | 177 |
> +-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+----------------+-----+
> 3 rows selected (0.056 seconds)
> {code}
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
