On Fri, Feb 20, 2015 at 2:22 PM, Dave Taht <dave.t...@gmail.com> wrote: > On Fri, Feb 20, 2015 at 12:33 PM, Juliusz Chroboczek > <j...@pps.univ-paris-diderot.fr> wrote: >>> I'd be a bit curious to know what people are using for test hardware. >> >> The WNDR3800/WNDR3700v2 is still my favourite. I've still got a couple >> Asus 500GP v1, and they're just fine if you don't need 802.11n and can >> fit in the slightly more limited flash. > > Despite evaluating well over 60 products and chipsets in the past 3 > years, I have yet to find something as nice as the wndr3800 as either > a R&D or day to day ultra-reliable home router. > > More generically the ag71xx/ath9k chipsets it used are the most > aimiable to hacking on and there are hundreds upon hundreds of current > products based on that, like the WZR the homewrt folk used. > > However single cpus like that with 32k dcache proved not up to the job > of 802.11ac and there is a lot of chaos and crap gear going around on > that standard. > > Elsewhere the whole wifi industry is seemingly in a race to the > bottom. As one example, I use the nanostation M5s heavily, and they > (silently) changed the underlying chipset last year from one with two > ethernet interfaces, to 1 connected to a 100Mbit switch. Result - no > link status anymore, the impossibility of measuring traffic well > across the two interfaces, and an increase of latency under load to > the buffering in the switch. > > The new nanostation IS a bit faster (mips 74k based rather than 24k), > and has twice the ram, for the same cost as the old, and it still has > 8MB flash, so I should not complain too much, I guess. (My typical > topology is 2 nanostations connected to a picostation M2HP AP) > > I have been seeking an integrated outdoor tri radio replacement for > that particular setup for a long time. > > As for everything else I have evaluated, I am thinking of publishing > it all, after I tone down the invective logged in my notes. > Most recently I took a look at two 802.11ac products from d-link. > > https://lists.bufferbloat.net/pipermail/bloat/2015-February/002310.html > > ubnt is *clued* in comparison to nearly everyone else.
The tp-link archer C7 v2, after a year of teething pains on the ath10k interface, is a not horrible platform for openwrt development (and the ath9k also on board, is great). Just dont expect the claimed 802.11ac rates on the wifi, it does not have enough cpu to push it that hard. It was only getting wndr3800 style forwarding rates when I last tried on on ethernet also. I was fairly impressed with the capabilities of the netgear x4, but was able to crash it regularly in testing when I last pulled it out and I dont think the gpl drop landed yet so there is no open source for it. There are a couple other 802.11ac routers out there like asus´s and so on, I do hope the market starts sorting out one or more that is useful to hack on. Felix recommended trying the d-link DIR-860L *model B*, but as yet I havent sorted out how to actually order that. He is also hot on the mt72 chipset in it as a candidate for multi-station queueing. I note that I have not, for most of the last year, been heads down into any of this stuff, I burned out pretty badly as we stablized cerowrt by last april and needed a major break. I happen to like gfiber´s new ethernet/wifi/video gw quite a bit, but that is only just now beginning be deployed in austin, and there is still a need for a followup release of the software, and I cant talk about it much, since I worked on it. I dont see much hope of seeing general support for the chipset it used (mindspeed) in mainline linux, either. .. I think that is all of my invective-free notes on various existing products. >> Both models: >> >> - are well supported by OpenWRT; >> - are very difficult to brick (just set up a tftp server and interrupt >> the boot sequence to reflash); >> - have a manageable built-in switch that's configurable from OpenWRT. >> >> The WNDR3800 has 16/128 flash/RAM, the 3700v2 16/64, the Asus 8/32. >> >> The WNDR3800/3700v2 has some other nice features, like a second gigabit >> NIC that doesn't go through the internal switch (to get VLANs without >> messing with the switch), support for 5GHz, and two radios usable >> simultaneously (which Babel is able to use for avoiding intra-route >> interference). > > With the current chaos calmer code (using linux 3.18) I get a > downstream rate of 560mbits and an upstream of 146 on the ethernet. > > The ratio used to be closer to 340/180. > > So much else in my lab (notably the tcps) has changed since the last > time I did benchmarking that I have no idea what to point at, and > openwrt enables the wndr3800 vlan by default, which is not my fav > thing, either, and I haven´t poked into any of it much, being > presently focused on evaluating the latest patchset for minstrel-blues > ( http://www.linuxplumbersconf.org/2014/ocw/sessions/2439 ) right now. > > >> -- Juliusz >> >> _______________________________________________ >> homenet mailing list >> homenet@ietf.org >> https://www.ietf.org/mailman/listinfo/homenet > > > > -- > Dave Täht > > thttp://www.bufferbloat.net/projects/bloat/wiki/Upcoming_Talks -- Dave Täht thttp://www.bufferbloat.net/projects/bloat/wiki/Upcoming_Talks _______________________________________________ homenet mailing list homenet@ietf.org https://www.ietf.org/mailman/listinfo/homenet
