[Spark SQL][reopen SPARK-16951]:Alternative implementation of NOT IN to Anti-join

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

[Spark SQL][reopen SPARK-16951]:Alternative implementation of NOT IN to Anti-join

Shuang, Linna1

Hello,

 

This JIRA (SPARK-16951) already being closed with the resolution of “Won’t Fix” on 23/Feb/17.

 

But in TPC-H test, we met performance issue of Q16, which used NOT IN subquery and being translated into broadcast nested loop join. This query uses almost half time of total 22 queries. For example, 512GB data set, totally execution time is 1400 seconds, while Q16’s execution time is 630 seconds.

 

TPC-H is a common spark sql performance benchmark, this performance issue will be met usually. Is it possible to reopen this JIRA and fix this issue?

 

Thanks,

Linna

 

Reply | Threaded
Open this post in threaded view
|

Re: [Spark SQL][reopen SPARK-16951]:Alternative implementation of NOT IN to Anti-join

Mich Talebzadeh
Hi Linna,

Please provide a background to it and your solution. The assumption is that there is a solution. as suggested.

Thanks,

Dr Mich Talebzadeh

 

LinkedIn  https://www.linkedin.com/profile/view?id=AAEAAAAWh2gBxianrbJd6zP6AcPCCdOABUrV8Pw

 

http://talebzadehmich.wordpress.com


Disclaimer: Use it at your own risk. Any and all responsibility for any loss, damage or destruction of data or any other property which may arise from relying on this email's technical content is explicitly disclaimed. The author will in no case be liable for any monetary damages arising from such loss, damage or destruction.

 



On Tue, 12 May 2020 at 04:47, Shuang, Linna1 <[hidden email]> wrote:

Hello,

 

This JIRA (SPARK-16951) already being closed with the resolution of “Won’t Fix” on 23/Feb/17.

 

But in TPC-H test, we met performance issue of Q16, which used NOT IN subquery and being translated into broadcast nested loop join. This query uses almost half time of total 22 queries. For example, 512GB data set, totally execution time is 1400 seconds, while Q16’s execution time is 630 seconds.

 

TPC-H is a common spark sql performance benchmark, this performance issue will be met usually. Is it possible to reopen this JIRA and fix this issue?

 

Thanks,

Linna

 

Reply | Threaded
Open this post in threaded view
|

RE: [Spark SQL][reopen SPARK-16951]:Alternative implementation of NOT IN to Anti-join

Shuang, Linna1

Hi Talebzadeh,

 

Thank you reply, the background is to use a common benchmark(here we use TPC-H) to compare different platform’s performance.

 

Our current solution is

  1. remove Q16 out of test
  2. rewrite Q16 without using “NOT IN”

 

Both solutions are not perfect. For Solution b which is suggested by the JIRA, the common benchmark already being changed, sometimes the result will be questioned when comparing with other platforms. In my understanding, the best solution is to find a better way to support “NOT IN” in Spark SQL, instead of suggesting not to use “NOT IN” in query.

 

Thanks,

Linna

 

From: Mich Talebzadeh <[hidden email]>
Sent: Tuesday, May 12, 2020 11:16 PM
To: Shuang, Linna1 <[hidden email]>
Cc: [hidden email]
Subject: Re: [Spark SQL][reopen SPARK-16951]:Alternative implementation of NOT IN to Anti-join

 

Hi Linna,

 

Please provide a background to it and your solution. The assumption is that there is a solution. as suggested.

 

Thanks,


Dr Mich Talebzadeh

 

LinkedIn  https://www.linkedin.com/profile/view?id=AAEAAAAWh2gBxianrbJd6zP6AcPCCdOABUrV8Pw

 

http://talebzadehmich.wordpress.com

 

Disclaimer: Use it at your own risk. Any and all responsibility for any loss, damage or destruction of data or any other property which may arise from relying on this email's technical content is explicitly disclaimed. The author will in no case be liable for any monetary damages arising from such loss, damage or destruction.

 

 

 

On Tue, 12 May 2020 at 04:47, Shuang, Linna1 <[hidden email]> wrote:

Hello,

 

This JIRA (SPARK-16951) already being closed with the resolution of “Won’t Fix” on 23/Feb/17.

 

But in TPC-H test, we met performance issue of Q16, which used NOT IN subquery and being translated into broadcast nested loop join. This query uses almost half time of total 22 queries. For example, 512GB data set, totally execution time is 1400 seconds, while Q16’s execution time is 630 seconds.

 

TPC-H is a common spark sql performance benchmark, this performance issue will be met usually. Is it possible to reopen this JIRA and fix this issue?

 

Thanks,

Linna