From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-ie0-x22e.google.com (mail-ie0-x22e.google.com [IPv6:2607:f8b0:4001:c03::22e]) (using TLSv1 with cipher RC4-SHA (128/128 bits)) (Client CN "smtp.gmail.com", Issuer "Google Internet Authority" (verified OK)) by huchra.bufferbloat.net (Postfix) with ESMTPS id 6CBFC21F0D3 for ; Sun, 26 May 2013 05:08:56 -0700 (PDT) Received: by mail-ie0-f174.google.com with SMTP id 10so15745392ied.5 for ; Sun, 26 May 2013 05:08:55 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20120113; h=mime-version:date:message-id:subject:from:to:content-type; bh=+MQ2Q8M4436/quD/casCLZlP0XoBUoMgSpQ8JeFwduE=; b=diCaqbEomk0jhBFEA9x79HoK9B2QRNCGX83E376PeLCEVXg+oElAnh2hRoTLw+5MS9 ecPHrkRDzxSyxPwnXIRVpLtlFYYEKRRoybKdEWq08vTFqlrPfwo/Rp9gdl97n2St2slE ibSp62ygZE8Vz+HCpOcteSwMz/AUmhMjcq0OHmp1flcFiCDXXru4UuaE3wLheufbS0AA 4epia0iMcUbEi+s8VyU7Rbz8nFzHbUd+tuCjJTwtCQkb83JB2GAXKVBb4C+UQDGNXvSD UySm0fyBC2RckL7T+A0NpENP25vzbb69aI4gnDTlcG27RHnguLP/wL4bH0phl+iadfgD kvyg== MIME-Version: 1.0 X-Received: by 10.42.54.129 with SMTP id r1mr15635645icg.23.1369570135020; Sun, 26 May 2013 05:08:55 -0700 (PDT) Received: by 10.64.35.44 with HTTP; Sun, 26 May 2013 05:08:54 -0700 (PDT) Date: Sun, 26 May 2013 05:08:54 -0700 Message-ID: From: Dave Taht To: cerowrt-devel@lists.bufferbloat.net, Felix Fietkau , Steven Barth Content-Type: multipart/alternative; boundary=90e6ba614b7a649a2304dd9de538 Subject: [Cerowrt-devel] tp-link 4300 evaluation X-BeenThere: cerowrt-devel@lists.bufferbloat.net X-Mailman-Version: 2.1.13 Precedence: list List-Id: Development issues regarding the cerowrt test router project List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sun, 26 May 2013 12:08:56 -0000 --90e6ba614b7a649a2304dd9de538 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable I am still on a quest to find a suitable netgear 3800 replacement. Paul Vixie found the same chipset (I hope!) being marketed as the WNDR3800-100NAS http://www.newegg.com/Product/Product.aspx?Item=3DN82E16833122434 so I have less fear about having to make the switch. That said there are a few products on the market from buffalo and tplink that are readily available on shelves today that seemed to be worth trying. esr and I picked up a tp-link 4300 (89 retail) for evaluation yesterday. At first glance it looked like a substantial update to the wndr3800 as it has a two generation leap forward on the chipset, using the Atheros AR9344 rev 2, which features the MIPs 74k processor, which supposedly is twice as fast as the 24k in the wndr3800. The box runs at a lower clock rate (540Mhz) than the netgear 3800 (680), but I figured the better processor would win, even if it got most of its theoretical win from an overly deep cpu pipeline= . I built openwrt for it, as I haven't built pure openwrt in a while... and this device only has 8MB of flash, and I didn't feel like stripping down cerowrt at the time. + There is a new skin for the luci gui which is quite nice, I don't know why I haven't been using it. + It was neat to see the ipv6 relay thing work + Procd worked for the first time (I've been trying to switch to this in cero for a while) - Then I started poking into the forwarding performance... Cerowrt regularly achieves 300+Mbit of ethernet forwarding performance in netperf. the tplink barely gets 130. the marketing for the tplink calls it a "n750" adding up the max theoretical values for the wifi channels, but I doubt, given what I get on ethernet, that it even comes close to that in reality. It turns out on the tplink there is only one ethernet chip, which is vlan-ed from the internal switch to the external port, instead of the two that are in the wndr3800. Now, there are quite a few differences between the software loads of openwrt and of cerowrt - notably the openwrt build is bridged to the wifi (I turned off the bridge, no difference in performance), and the default atheros build currently has some de-optimizations in it to handle unaligned access to packets that I don't think are needed on this later chipset, but... it currently looks like in the quest for cost reduction, this step forward to "modern" hardware is actually a large step backwards. --=20 Dave T=E4ht Fixing bufferbloat with cerowrt: http://www.teklibre.com/cerowrt/subscribe.html --90e6ba614b7a649a2304dd9de538 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable I am still on a quest to find a suitable netgear 3800 replacement. Paul Vix= ie found the same chipset (I hope!)
being marketed as the WNDR3800-100NA= S http://www.newegg.com/Product/Product.aspx?Item=3DN82E16833122434<= br> so I have less fear about having to make the switch.

That said there= are a few products on the market from buffalo and tplink that are readily = available on shelves today that seemed to be worth trying.

esr and I= picked up a tp-link 4300 (89 retail) for evaluation yesterday. At first gl= ance it looked like a substantial update to the wndr3800 as it has a two ge= neration leap forward on the chipset, using the Atheros AR9344 rev 2, which= features the MIPs 74k processor, which supposedly is twice as fast as the = 24k in the wndr3800. The box runs at a lower clock rate (540Mhz) than the n= etgear 3800 (680), but I figured the better processor would win, even if it= got most of its theoretical win from an overly deep cpu pipeline.

I built openwrt for it, as I haven't built pure openwrt in a while.= .. and this device only has 8MB of flash, and I didn't
feel like str= ipping down cerowrt at the time.

+ There is a new skin for the luci= gui which is quite nice, I don't know why I haven't been using it.=
+ It was neat to see the ipv6 relay thing work
+ Procd worked for the fi= rst time (I've been trying to switch to this in cero for a while)
- = Then I started poking into the forwarding performance...

Cerowrt regularly achieves 300+Mbit of ethernet forwarding performance = in netperf. the tplink barely gets 130. the marketing for the tplink calls = it a "n750" adding up the max theoretical values for the wifi cha= nnels, but I doubt, given what I get on ethernet, that it even comes close = to that in reality.

It turns out on the tplink there is only one ethernet chip, which is vl= an-ed from the internal switch to the external port, instead of the two tha= t are in the wndr3800.

Now, there are quite a few differences betwee= n the software loads of openwrt and of cerowrt - notably the openwrt build = is bridged to the wifi (I turned off the bridge, no difference in performan= ce), and the default atheros build currently has some de-optimizations in i= t to handle unaligned access to packets that I don't think are needed o= n this later chipset, but... it currently looks like in the quest for cost = reduction, this step forward to "modern" hardware is actually a l= arge step backwards.

--
Dave T=E4ht

Fixing bufferbloat with cerowrt: http://www= .teklibre.com/cerowrt/subscribe.html=20 --90e6ba614b7a649a2304dd9de538--