From mboxrd@z Thu Jan 1 00:00:00 1970 From: Dario Faggioli Subject: Re: [PATCH 00/24] sched: Credit1 and Credit2 improvements... and soft-affinity for Credit2! Date: Thu, 18 Aug 2016 02:11:01 +0200 Message-ID: <1471479061.6806.54.camel@citrix.com> References: <147145358844.25877.7490417583264534196.stgit@Solace.fritz.box> Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="===============3817477160704276288==" Return-path: Received: from mail6.bemta6.messagelabs.com ([193.109.254.103]) by lists.xenproject.org with esmtp (Exim 4.84_2) (envelope-from ) id 1baAvN-0002ls-Ct for xen-devel@lists.xenproject.org; Thu, 18 Aug 2016 00:11:09 +0000 In-Reply-To: <147145358844.25877.7490417583264534196.stgit@Solace.fritz.box> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Errors-To: xen-devel-bounces@lists.xen.org Sender: "Xen-devel" To: xen-devel@lists.xenproject.org Cc: Wei Liu , Andrew Cooper , Anshul Makkar , Ian Jackson , George Dunlap , David Vrabel , Jan Beulich List-Id: xen-devel@lists.xenproject.org --===============3817477160704276288== Content-Type: multipart/signed; micalg=pgp-sha256; protocol="application/pgp-signature"; boundary="=-KVBtRmtPfseIq88uvHwl" --=-KVBtRmtPfseIq88uvHwl Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable On Wed, 2016-08-17 at 19:17 +0200, Dario Faggioli wrote: > The last 4 patches, still for Credit2, are optimizations, either wrt > existing > code, or wrt new code introduced in this series. I've chosen to keep > them > separate to make reviewing/understanding new code easier. In fact, > although > they look pretty simple, the soft-affinity code was pretty complex > already, and > even these simple optimization, if done all at once, would have made > the > reviewer's life (unnecessary) tougher. >=20 About this. I've run the benchmarks with and without these performance optimization patches, in order to assess their effect as good as I could. The baseline on top of which I was applying the series is different from the one used to produce the other numbers reported in the cover letter, so what's shown there and what I show here is not directly comparable (but that's not a problem). Given the nature of the improvements, I've run more iterations of each configuration of the benchmarks (i.e., 15 iterations instead of 5) to get more stable results. Here's my findings: ++++++++++++++++++++++++++++++++++ | =C2=A0 =C2=A0CREDIT1, for reference =C2=A0 =C2=A0 =C2=A0| ++++++++++++++++++++++++++++++++++ | =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 | MAKEXEN IPERF =C2=A0| |---------------|----------------| |no dom0 load =C2=A0 | 28.353 =C2=A011.793 | |with dom0 load | 43.955 =C2=A010.932*| ++++++++++++++++++++++++++++++++++ ++++++++++++++++++++++++++++++++++ | =C2=A0 CREDIT2, until patch 20 =C2=A0 =C2=A0 =C2=A0| ++++++++++++++++++++++++++++++++++ | =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 | MAKEXEN IPERF =C2=A0| |---------------| ---------------| |no dom0 load =C2=A0 | 28.367 =C2=A011.716 | |with dom0 load | 40.591 =C2=A010.645 | ++++++++++++++++++++++++++++++++++ |+++++++++++++++++++++++++++++++++ | =C2=A0 =C2=A0CREDIT2, full series =C2=A0 =C2=A0 =C2=A0 =C2=A0| ++++++++++++++++++++++++++++++++++ | =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 | MAKEXEN IPERF =C2=A0| |---------------|----------------| |no dom0 load =C2=A0 | 27.597* 12.059*| |with dom0 load | 39.706* 10.609 | |--------------------------------| =C2=A0* marks the best results So: =C2=A0- apart from a glitch on "IPERF with dom0 load", Credit2 with the=C2= =A0 =C2=A0 =C2=A0full series applied is confirmed to be the best. About the gli= tch: =C2=A0 =C2=A0 - wrt the fact that Credit1 is better, we also have other evi= dences =C2=A0 =C2=A0 =C2=A0 that network throughput could be a bit of a weak spot = of Credit2 =C2=A0 =C2=A0 =C2=A0 versus Credit1 so far (although, we have to admit, the= y're =C2=A0 =C2=A0 =C2=A0 pretty close), and we already have ideas on how to try= improve =C2=A0 =C2=A0 =C2=A0 the situation; =C2=A0 =C2=A0 - wrt the role played by optimization patches, well, results = are =C2=A0 =C2=A0 =C2=A0 basically the same. =C2=A0- The performance optimization patches do have an (positive!) impact. =C2=A0- In case of "no dom0 load, it's actually thanks to the optimization =C2=A0 =C2=A0patches that Credit2 beats Credit1. Regards, Dario --=20 <> (Raistlin Majere) ----------------------------------------------------------------- Dario Faggioli, Ph.D, http://about.me/dario.faggioli Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK) --=-KVBtRmtPfseIq88uvHwl Content-Type: application/pgp-signature; name="signature.asc" Content-Description: This is a digitally signed message part Content-Transfer-Encoding: 7bit -----BEGIN PGP SIGNATURE----- Version: GnuPG v2 iQIcBAABCAAGBQJXtP0VAAoJEBZCeImluHPu+QgP/15kj6TnwRFfKmhmqZTRxric JLpxBzZ1Iv1hDXLt8ZGwT/B+41/QEXqf4hX4KWgt/dc8oEXoUcIDsi6uGrTwMp3M L2tbFTWxBKre5HI2A0Lztmqh5HnpdswsRfY2BmYBAaCKOtKKo7NPwbRlE+ONR0wE fwcTieCDh1ZgpzfYCKeKLMDA9Kpda5AoRNyCHSuSb6bClCuA+l4gur263UOsY4CH Cyv8UxG+8lrOOk7mDkKlyXG9jgJSZEu7ZDBWLW2B55+aQNPC3NsyXZy0tS23V9b5 FAVpm1GxjGhkZGNaW5nWe2OCIE4gmxgeLy7P24pnect1FlRzb8R9ne1AzsSNYSDN 22DnoE1bSYQZhivjrntuSwGgz0y4FAxpHzwHqcsBE3OcqDQegs3qVrslSn4oCVwE phWYzNy2eOrRl02hTYeWUmB5sLbjolvJImulwRuIW9rG6OC/S9NFJ+OtgtOH6N1F 0s5ai+qNi4mcf6PQwUHPigWmmG741VcSoTdPsKlg1IuhZSJTBNY9WHzHP3dhceGk phs62SfFUHFZ9x3+y4uuwemgLsVpwJGnmWhozco7Qs9at+NCdvTLGdZQLPuU+0ty Tvwj0Ez8GXjp1FT8ERhhVR9umEiWv/IBTLY2IJVw6MXiJYN/her3+b5H3RGQz0H3 hpQrcyI8bRlGp76AkxtL =Oo1E -----END PGP SIGNATURE----- --=-KVBtRmtPfseIq88uvHwl-- --===============3817477160704276288== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: base64 Content-Disposition: inline X19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX18KWGVuLWRldmVs IG1haWxpbmcgbGlzdApYZW4tZGV2ZWxAbGlzdHMueGVuLm9yZwpodHRwczovL2xpc3RzLnhlbi5v cmcveGVuLWRldmVsCg== --===============3817477160704276288==--