[Starlingx-discuss] How low latency implemented in STX?

Jones, Bruce E bruce.e.jones at intel.com
Thu Feb 18 17:40:21 UTC 2021


Forgive my top-posting email client.

Yes, you can deploy the low latency profile for worker nodes and get the rt kernel.

No, I’m not aware of anything in StarlingX specific to video streaming.  I wouldn’t even know what’s needed, perhaps there are folks in the community who have that background.

        brucej

From: open infra <openinfradn at gmail.com>
Sent: Thursday, February 18, 2021 7:51 AM
To: Jones, Bruce E <bruce.e.jones at intel.com>
Cc: starlingx-discuss at lists.starlingx.io
Subject: Re: [Starlingx-discuss] How low latency implemented in STX?


Hi
On Thu, Feb 18, 2021 at 8:50 PM Jones, Bruce E <bruce.e.jones at intel.com<mailto:bruce.e.jones at intel.com>> wrote:
There are several low latency features in StarlingX, depending on what you mean by low latency.

I just asked to understand as 'ultra-low latency'  feature is mentioned in the stx web site.


StarlingX supports the rt-linux kernel, which significantly improves kernel scheduling and reduces rt application latency.

So, we use low latency profile with RT kernel?
Or customized profile that specific to stx?


It also supports OVS DPDK, which greatly reduces network packet delivery latency, which I believe is only supported today in Neutron.  There is work in progress in the communitfeature y to enable DPDK in Kubernetes but AFAIK that work has not yet been picked up by StarlingX.

Thanks for sharing this information.

Could be a dumb question:
Does it provide any specific protocols to improve the performance of video streaming etc?


     brucej

From: open infra <openinfradn at gmail.com<mailto:openinfradn at gmail.com>>
Sent: Thursday, February 18, 2021 7:12 AM
To: starlingx-discuss at lists.starlingx.io<mailto:starlingx-discuss at lists.starlingx.io>
Subject: [Starlingx-discuss] How low latency implemented in STX?

Hi,


I would like to know how low latency implemented in StarlingX.
- Is it implemented only in starlingx itself
or
Is it implemented in Kubernetes shipped with starlingx?
Any optimization has been done to Neutron (under OpenStack) in order to achieve low latency with virtualized networking?
or do we use exact same upstream OpenStack components under StarlingX?

Regards,
Danishka

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.starlingx.io/pipermail/starlingx-discuss/attachments/20210218/81db60fb/attachment-0001.html>


More information about the Starlingx-discuss mailing list