why default value of hawq_rm_nvseg_perquery_perseg_limit is 6?

[Guo Kai <gu...@gmail.com>]

Hi, all!

I am curious why default value of hawq_rm_nvseg_perquery_perseg_limit is 6.
Is it an empirical value? Or does it exist a constrain on a vseg's workload
or computing power?

Thanks!

-- 
Guo Kai,
guokaiwhu@gmail.com

Re: why default value of hawq_rm_nvseg_perquery_perseg_limit is 6?

[Vineet Goel <vv...@apache.org>]

Guo,

This is mostly based on empirical data that's a safe "default value" and
provides a good balance between parallelism and concurrency for typical
environments. If you have more resources (Memory/CPU) on your cluster, you
could certainly move the value up to 8 or 10+ and test the performance on
your workloads and user needs. There are multiple variables that go into
tuning your cluster which depend on resources, concurrency, type of
workload (query plans generated) ...etc.

Hope this helps.

-Vineet



On Fri, Mar 17, 2017 at 7:19 AM Guo Kai <gu...@gmail.com> wrote:

> Hi, all!
>
> I am curious why default value of hawq_rm_nvseg_perquery_perseg_limit is
> 6. Is it an empirical value? Or does it exist a constrain on a vseg's
> workload or computing power?
>
> Thanks!
>
> --
> Guo Kai,
> guokaiwhu@gmail.com
>