From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-oi0-x22f.google.com (mail-oi0-x22f.google.com [IPv6:2607:f8b0:4003:c06::22f]) (using TLSv1 with cipher RC4-SHA (128/128 bits)) (Client CN "smtp.gmail.com", Issuer "Google Internet Authority G2" (verified OK)) by huchra.bufferbloat.net (Postfix) with ESMTPS id EF47A21F38B for ; Mon, 23 Feb 2015 11:03:49 -0800 (PST) Received: by mail-oi0-f47.google.com with SMTP id i138so15226948oig.6 for ; Mon, 23 Feb 2015 11:03:48 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20120113; h=mime-version:in-reply-to:references:date:message-id:subject:from:to :cc:content-type; bh=yoR+TBkcv7oKBmVBZKUjTwQpHsevqMg8eLRM4g5Szu0=; b=A3wviywu3fT9ZtYnumaRGyhaS4ZNL8N484pQHY/0doCQKERRX8M2haYTx4c39Fr/Ib qd0SnUB2x6dQm7iuzXm2pKGo7aZqkarRqAD/m0h5BI+TiDWcAcCKixufQBtj/Bb0HuXD 5eF2bDMwJIzdencGexRmtJ99f4ULdbTl5cyszF4M9EayM20/ajnGLsa9sEYH2p7vKaMJ FNeFs25t7sV+5LHY+WNAyZlg4Wj77ETq8gQEoLL4Oo/cphox9Uk9h7v5B3qyNde/wILv XnAmOXcIU2NQPVm+RKvgTPADWSgH82BZZ+yRCd6ua+6xAFA8x38wWfSFGFxzAeqlWc3q a7Mw== MIME-Version: 1.0 X-Received: by 10.60.16.230 with SMTP id j6mr8510610oed.8.1424718228661; Mon, 23 Feb 2015 11:03:48 -0800 (PST) Received: by 10.202.51.66 with HTTP; Mon, 23 Feb 2015 11:03:48 -0800 (PST) In-Reply-To: References: Date: Mon, 23 Feb 2015 11:03:48 -0800 Message-ID: From: Dave Taht To: "Bill Ver Steeg (versteb)" Content-Type: multipart/alternative; boundary=e89a8f503bb2ecacc2050fc60fe8 Cc: "Bloat@lists.bufferbloat.net" Subject: Re: [Bloat] recommended PC config for network testing using Ubuntu X-BeenThere: bloat@lists.bufferbloat.net X-Mailman-Version: 2.1.13 Precedence: list List-Id: General list for discussing Bufferbloat List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 23 Feb 2015 19:04:18 -0000 --e89a8f503bb2ecacc2050fc60fe8 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: quoted-printable I was a little unclear - I use the nucs as a my primary linux desktop in 3 different locations, and they are also test clients on the networks I am testing. They *just work* with all the kernels I build, and the video driver, in particular, is excellent for day to day use, even with multiple hires displays. You can bolt them on the back of a monitor that supports that, getting rid of unsightly cables. I am always finding a need to have a 5th usb port, but that is me.... The shorter ones are good if all you need is a half high wifi card and use an msata hard disk, the taller ones with a sata SSD are way faster and let you use a full length wifi card. I do sometimes regret not having got the i5 or better ones, but I offload major compilations into googles cloud and snapon - or just toss them into the background - so the speed rarely bothers me on my day to day workloads. If I were a gamer I would consider something else. I run ubuntu gnome (various versions) rather than normal ubuntu as I hate their present gui direction. While I am dumping my stack about hardware I am huge fan of buckling spring keyboards http://www.pckeyboard.com/page/FeaturedProducts/UB40PGA They have the same nubbie thing that the lenovo laptops have for a mouse pointer, they are loud enough to annoy co-workers across the room, and they let me type accurately at insane speeds and flood peoples mailboxes with email. :) The vast majority of the other working parts of my lab are wndr3800, ubnt picostation, and ubnt nanostation. The beaglebones mostly did not work out, nor the rasberry pi - too easy to roach the filesystems. I have a ton of other embedded hardware that didnt work out worse, but to talk about them requires editing out a lot of epithets. Avoid the globalscale products like the plague they are in particular - they have ancient kernels and run way too hot. As already noted, most of the 32 bit arm chips have lousy ethernet - and most lousy video drivers. The arm64 stuff is starting to work but havent touched it in a while. People keep asking me to try out the wandboard, haven=C2=B4t. I have a zedboard. no bql, but easy to add. there is really promising work going on around the zynq FPGA in particular. These guys might be onto something, but they havent returned my mail with questions about their "35" product. https://www.kickstarter.com/projects/onetswitch/onetswitch-open-source-hard= ware-for-networking lastly, I meant to include a plot of rangeley=C2=B4s behavior: http://snapon.lab.bufferbloat.net/~d/rangeley/fq2fq_vs_pfifo_fast_rangeley.= png it seems impossible to get a modern linux architecture under load down much below 2ms at gigE at least in part due to context switch overhead. I have a bit of hope for the dpdk work after the recent preso by stephen at netconf01.org /me stops ranting, goes back to work --e89a8f503bb2ecacc2050fc60fe8 Content-Type: text/html; charset=UTF-8 Content-Transfer-Encoding: quoted-printable
I was a little unclear - I use = the nucs as a my primary linux desktop in 3 different locations, and they a= re also test clients on the networks I am testing. They *just work* with al= l the kernels I build, and the video driver, in particular, is excellent fo= r day to day use, even with multiple hires displays. You can bolt them on t= he back of a monitor that supports =C2=A0that, getting rid of unsightly cab= les. I am always finding a need to have a 5th usb port, but that is me....

The s= horter ones are good if all you need is a half high wifi card and use an ms= ata hard disk, the taller ones with a sata SSD are way faster and let you u= se a full length wifi card.

I do sometimes regret not having got the i5 or better= ones, but I offload major compilations into googles cloud and snapon - or = just toss them into the background - so the speed rarely bothers me on my d= ay to day workloads.=C2=A0

If I were a gamer I would consider something else.

I run ubu= ntu gnome (various versions) rather than normal ubuntu as I hate their pres= ent gui direction.=C2=A0

While I am dumping my stack about hardware I am huge fan= of buckling spring keyboards

They have the same = nubbie thing that the lenovo laptops have for a mouse pointer, they are lou= d enough to annoy co-workers across the room, and they let me type accurate= ly at insane speeds and flood peoples mailboxes with email. =C2=A0:)
<= div class=3D"gmail_extra">
The vast maj= ority of the other working parts of my lab are wndr3800, ubnt picostation, = and ubnt nanostation. The beaglebones mostly did not work out, nor the rasb= erry pi - too easy to roach the filesystems. I have a ton of other embedded= hardware that didnt work out worse, but to talk about them requires editin= g out a lot of epithets. Avoid the globalscale products like the plague the= y are in particular - they have ancient kernels and run way too hot. As alr= eady noted, most of the 32 bit arm chips have lousy ethernet - and most lou= sy video drivers. The arm64 stuff is starting to work but havent touched it= in a while.

People keep asking me to try out the wandboard, haven=C2=B4t.
<= div class=3D"gmail_extra">
I have a zed= board. no bql, but easy to add. there is really promising work going on aro= und the zynq FPGA in particular. These guys might be onto something, but th= ey havent returned my mail with questions about their "35" produc= t.=C2=A0


lastly, I meant to= include a plot of rangeley=C2=B4s behavior:


it seems impos= sible to get a modern linux architecture under load down much below 2ms at = gigE at least in part due to context switch overhead. I have a bit of hope = for the dpdk work after the recent preso by stephen at netconf01.org

/me stops ranting, goes back to work



--e89a8f503bb2ecacc2050fc60fe8--