Colin Ian King
colin.king at canonical.com
Tue Jul 27 21:43:57 UTC 2021
On 27/07/2021 19:57, Tim Gardner wrote:
> Does anyone have experience with this config option ?
I did some analysis a while ago (31st March 2020) and Cc'd Francis and
Seth on the results. Here is what I wrote:
"I ran some synthetic stress-ng throughput and latency tests with and
without CONFIG_RT_GROUP_SCHED on CONFIG_RT_GROUP_SCHED. I don't see
enough variation in the throughput or latency to warrant any concern
about enabling this option. We have quite a bit of small jitter on the
results, but overall the kernel results are similar enough to make me
feel OK to enable this option for Focal.
Attached is a LibreOffice spread sheet with my results. I didn't run
the LKP tests as these take ~15 hours per run to complete and I don't
think it will give us much more useful information."
>It is my
> impression that without allocating any RT threads that there is no
> change in default behavior. A brief survey of code enabled by
> CONFIG_RT_GROUP_SCHED _kind_ of reinforces that observation. The reason
> I ask is that Microsoft has been testing this config option with the
> Azure kernel. It has satisfied at least one customer that wanted this
> option. They (MSFT) plan to test to make sure its really doing what they
> want as well as watching for behavioral changes or regressions. I
> thought adding any information from y'all would be helpful.
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 20654 bytes
Desc: not available
More information about the kernel-team