DPDK patches and discussions
 help / color / mirror / Atom feed
From: "Dumitrescu, Cristian" <cristian.dumitrescu@intel.com>
To: Sridhar.V.Iyer <sridhariyer@versa-networks.com>,
	"dev@dpdk.org" <dev@dpdk.org>
Cc: Ananda <ananda@versa-networks.com>
Subject: Re: [dpdk-dev] Old oversubscription related checkin
Date: Thu, 5 May 2016 12:49:52 +0000	[thread overview]
Message-ID: <3EB4FA525960D640B5BDFFD6A3D89126479B78EA@IRSMSX108.ger.corp.intel.com> (raw)
In-Reply-To: <140ABA5F-E537-409E-8614-D3B986C6121C@versa-networks.com>

Hi Sridhar,

I think this patch is simply the implementation of the oversubscription feature, as introduced by DPDK release 1.4.1. Most likely, this patch is the way Thomas added this Intel release to dpdk.org.

The oversubscription feature was never implemented for all 4 traffic classes. The patch you are pointing to simply introduced this feature for TC3 only. Before this patch, there might have been just a placeholder in the code for the oversubscription feature that was to be developed later on.

Again, the reason for introducing this feature for TC3 only is that the higher priority traffic classes TC0 .. TC2 are typically fully provisioned, as the amount of TC0 .. TC2 traffic is much less than TC3 (Best Effort) traffic, which is usually hugely overprovisioned. This feature can potentially be extended to TC0 .. TC2 as well, as the way this feature basically works is deciding the quanta to be applied for all pipes in the subport during the next scheduling decision based on the amount of unused credits found at the end of the current scheduling period.

Regards,
Cristian

From: Sridhar.V.Iyer [mailto:sridhariyer@versa-networks.com]
Sent: Wednesday, May 4, 2016 10:15 PM
To: dev@dpdk.org; Dumitrescu, Cristian <cristian.dumitrescu@intel.com>
Cc: Ananda <ananda@versa-networks.com>
Subject: Old oversubscription related checkin

Hi Cristian,

 I stumbled into an old from 2013 (http://dpdk.org/browse/dpdk/patch/lib/librte_sched/rte_sched.c?id=835c5409a7bac3055b82bebee65d8ada7f20d332)

I couldn’t find any context for the patch from the mail archives. Could you please let me know why oversubscription was removed from all the other traffic classes and just given to tc3?
Were there huge performance penalties? Would there be any issues if I add this patch back again locally (enable/disabled via config).

Regards,

Sridhar V Iyer

<http://www.versa-networks.com/>






<http://www.versa-networks.com/>


<http://www.versa-networks.com/>


<http://www.versa-networks.com/>


      reply	other threads:[~2016-05-05 12:49 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-05-04 21:14 Sridhar.V.Iyer
2016-05-05 12:49 ` Dumitrescu, Cristian [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=3EB4FA525960D640B5BDFFD6A3D89126479B78EA@IRSMSX108.ger.corp.intel.com \
    --to=cristian.dumitrescu@intel.com \
    --cc=ananda@versa-networks.com \
    --cc=dev@dpdk.org \
    --cc=sridhariyer@versa-networks.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).